This ioctl exposes VOP_BMAP information to userland. It can be used by
programs like fragmentation analyzers and optimized cp implementations. But
I'm using it to test fusefs's VOP_BMAP implementation. The "2" in the name
distinguishes it from the similar but incompatible FIBMAP ioctls in NetBSD
and Linux. FIOBMAP2 differs from FIBMAP in that it uses a 64-bit block
number instead of 32-bit, and it also returns runp and runb.
Reviewed by: mckusick
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D20705
ioctl(2) commands only have meaning in the context of a file descriptor
so translating them in the syscall layer is incorrect.
The new handler users an accessor to retrieve/construct a pointer from
the last member of the passed structure and relies on type punning to
access the other member which requires no translation.
Unlike r339174 this change supports both places FIODGNAME is handled.
Reviewed by: kib
Obtained from: CheriBSD
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D17475
Mainly focus on files that use BSD 3-Clause license.
The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.
Special thanks to Wind River for providing access to "The Duke of
Highlander" tool: an older (2014) run over FreeBSD tree was useful as a
starting point.
This makes roundup2/rounddown2 type- and const-preserving and allows
using it on pointer types without casting to uintptr_t first. Not
performing pointer-to-integer conversions also helps the compiler's
optimization passes and can therefore result in better code generation.
When using it with integer values there should be no change other than
the compiler checking that the alignment value is a valid power-of-two.
I originally implemented these builtins for CHERI a few years ago and
they have been very useful for CheriBSD. However, they are also useful
for non-CHERI code so I was able to upstream them for Clang 10.0.
Rationale from the clang documentation:
Clang provides builtins to support checking and adjusting alignment
of pointers and integers. These builtins can be used to avoid relying
on implementation-defined behavior of arithmetic on integers derived
from pointers. Additionally, these builtins retain type information
and, unlike bitwise arithmetic, they can perform semantic checking on
the alignment value.
There is also a feature request for GCC, so GCC may also support it in
the future: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98641
Reviewed By: brooks, jhb, imp
Differential Revision: https://reviews.freebsd.org/D28332
The age of the intel compiler support is so old as to be
uninteresting. No recent recports of intel compiler support have been
received. Remove all the special case workarounds for the Intel
compiler. Should there be interest in supporting the compiler, contact
me and I'll work with people to make it happen, though I suspect these
instances are more likely to be in the way than to be helpful.
Reviewed by: cem, emaste, vangyzen, dim
Differential Revision: https://reviews.freebsd.org/D26817
This fixes the following warning in libbsd:
rtems/blkdev.h:200:10: warning: implicit declaration of function 'ioctl'; did
you mean 'ifioctl'? [-Wimplicit-function-declaration]
Remove unnecessary includes.
Define FD_SETSIZE (<sys/select.h>) to be 1024 by default, and define
NOFILE (<sys/param.h>) to be OPEN_MAX (== 3200) by default.
Remove the comment in <sys/select.h> that FD_SETSIZE should be >=
NOFILE.
Bump API minor.
Addresses: https://cygwin.com/pipermail/cygwin/2022-July/251839.html
This reverts commit 1f8f7e2d54, "* libc/stdio/refill.c (__srefill):
Try again after EOF on Cygwin." If EOF is set on a file, the stdio
input functions will now immediately return EOF rather than trying
again to read. This aligns Cygwin's behavior to that of Linux.
Addresses: https://cygwin.com/pipermail/cygwin/2022-June/251672.html
The introduction of <sched.h> improved compatibility with some 3rd
party software, but caused the configure scripts of some ports to
assume that they were run in a GLIBC compatible environment.
Parts of sched.h were made conditional on -D_WITH_CPU_SET_T being
added to ports, but there still were compatibility issues due to
invalid assumptions made in autoconfigure scripts.
The differences between the FreeBSD version of macros like CPU_AND,
CPU_OR, etc. and the GLIBC versions was in the number of arguments:
FreeBSD used a 2-address scheme (one source argument is also used as
the destination of the operation), while GLIBC uses a 3-adderess
scheme (2 source operands and a separately passed destination).
The GLIBC scheme provides a super-set of the functionality of the
FreeBSD macros, since it does not prevent passing the same variable
as source and destination arguments. In code that wanted to preserve
both source arguments, the FreeBSD macros required a temporary copy of
one of the source arguments.
This patch set allows to unconditionally provide functions and macros
expected by 3rd party software written for GLIBC based systems, but
breaks builds of externally maintained sources that use any of the
following macros: CPU_AND, CPU_ANDNOT, CPU_OR, CPU_XOR.
One contributed driver (contrib/ofed/libmlx5) has been patched to
support both the old and the new CPU_OR signatures. If this commit
is merged to -STABLE, the version test will have to be extended to
cover more ranges.
Ports that have added -D_WITH_CPU_SET_T to build on -CURRENT do
no longer require that option.
The FreeBSD version has been bumped to 1400046 to reflect this
incompatible change.
Reviewed by: kib
MFC after: 2 weeks
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D33451
The changes to the bitset macros allowed sched.h to be included
into userland programs without name space pollution due to BIT_*
and BITSET_* macros.
The definition of a "struct bitset" had been overlooked. This name
space pollution caused the build of port print/miktex to fail.
This commit makes the definition of struct bitset depend on the
same condition as the visibility of the BIT_* and BITSET_* macros,
i.e. needs _KERNEL or _WANT_FREEBSD_BITSET to be defined before
including sys/_bitset.h.
It has been tested with "make universe" since a prior attempt to
fix the issue broke the PowerPC64 kernel build.
This commit shall be MFCed together with commit 5e04571cf3c.
Reported by: arrowd
MFC after: 1 month
It caused kernel build for PowerPC64 to fail.
A different patch is being tested with make universe to make sure it
works on all architectures.
MFC after: 1 month<N [day[s]|week[s]|month[s]]. Request a reminder email>
There is a reference to malloc() in #define __BITSET_ALLOC. Even
though this macro is only defined but not used, it causes the lang/gcc
ports to fail. The gcc ports "poison" a number of functions including
malloc() and prevent their use (including in macro definitions).
This commit moved the declaration of __BITSET_ALLOC into the
conditional block that depends on _KERNEL or _WANT_FREEBSD_BITSET
being defined.
There is no use of __BITSET_ALLOC in the FreeBSD sources, and userland
programs that want to use BITSEC_ALLOC will define _WANT_FREEBSD_BITSET
anyway.
This patch has been tested by building lang/gcc11 and a successful
make buildworld.
This commit shall be MFCed together with commit 5e04571cf3c.
MFC after: 1 month
The changes to the bitset macros allowed sched.h to be included into
userland programs without name space pollution due to BIT_* and
BITSET_* macros.
The definition of a global variable "bitset" had been overlooked.
This name space pollution caused a compile failure in print/miktex.
This commit renames the bitset variable to __bitset with the same
mapping back to the bitset if _KERNEL or _WANT_FREEBSD_BITSET is
defined.
This fix has been suggested by kib. It has been tested to let the
build of the print/miktex port succeed and to not break buildworld.
This commit shall be MFCed together with commit 5e04571cf3c.
Reported by: arrowd
MFC after: 1 month
Add two underscore characters "__" to names of BIT_* and BITSET_*
macros to move them to the implementation name space and to prevent
a name space pollution due to BIT_* macros in 3rd party programs with
conflicting parameter signatures.
These prefixed macro names are used in kernel header files to define
macros in e.g. sched.h, sys/cpuset.h and sys/domainset.h.
If C programs are built with either -D_KERNEL (automatically passed
when building a kernel or kernel modules) or -D_WANT_FREENBSD_BITSET
(or this macros is defined in the source code before including the
bitset macros), then all macros are made visible with their previous
names, too. E.g., both __BIT_SET() and BIT_SET() are visible with
either of _KERNEL or _WANT_FREEBSD_BITSET defined.
The main reason for this change is that some 3rd party sources
including sched.h have been found to contain conflicting BIT_*
macros.
As a work-around, parts of shed.h have been made conditional and
depend on _WITH_CPU_SET_T being set when sched.h is included.
Ports that expect the full functionality provided by sched.h need
to be built with -D_WITH_CPU_SET_T. But this leads to conflicts if
BIT_* macros are defined in that program, too.
This patch set makes all of sched.h visible again without this
parameter being passed and without any name space pollution due
to BIT_* macros becoming visible when sched.h is included.
This patch set will be backported to the STABLE branches, but ports
will need to use -D_WITH_CPU_SET_T as long as there are supported
releases that do not contain these patches.
Reviewed by: kib, markj
MFC after: 1 month
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D33235
This implementation is faster and doesn't modify the cpuset, so it lets
us avoid some unnecessary copying as well. No functional change
intended.
This is a re-application of commit
9068f6ea697b1b28ad1326a4c7a9ba86f08b985e.
Reviewed by: cem, kib, jhb
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D32029
Eliminate the nested loops and re-implement following a suggestion from
rlibby.
Add some simple regression tests.
Reviewed by: rlibby, kib
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D32472
This reverts commit 9068f6ea697b1b28ad1326a4c7a9ba86f08b985e.
The underlying macro needs to be reworked to avoid problems with control
flow statements.
Reported by: rlibby
This implementation is faster and doesn't modify the cpuset, so it lets
us avoid some unnecessary copying as well. No functional change
intended.
Reviewed by: cem, kib, jhb
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D32029
These allow one to non-destructively iterate over the set or clear bits
in a bitset. The motivation is that we have several code fragments
which iterate over a CPU set like this:
while ((cpu = CPU_FFS(&cpus)) != 0) {
cpu--;
CPU_CLR(cpu, &cpus);
<do something>;
}
This is slow since CPU_FFS begins the search at the beginning of the
bitset each time. On amd64 and arm64, CPU sets have size 256, so there
are four limbs in the bitset and we do a lot of unnecessary scanning.
A second problem is that this is destructive, so code which needs to
preserve the original set has to make a copy. In particular, we have
quite a few functions which take a cpuset_t parameter by value, meaning
that each call has to copy the 32 byte cpuset_t.
The new macros address both problems.
Reviewed by: cem, kib
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D32028
iflib now supports mapping each (TX,RX) queue pair to the same CPU
(default), to separate CPUs, or to a pair of physical and logical CPUs
that share the same L2 cache. The mapping mechanism supports unequal
numbers of TX and RX queues, with the excess queues always being
mapped to consecutive physical CPUs. When the platform cannot
distinguish between physical and logical CPUs, all are treated as
physical CPUs. See the comment on get_cpuid_for_queue() for the
entire matrix.
The following device-specific tunables influence the mapping process:
dev.<device>.<unit>.iflib.core_offset (existing)
dev.<device>.<unit>.iflib.separate_txrx (existing)
dev.<device>.<unit>.iflib.use_logical_cores (new)
The following new, read-only sysctls provide visibility of the mapping
results:
dev.<device>.<unit>.iflib.{t,r}xq<n>.cpu
When an iflib driver allocates TX softirqs without providing reference
RX IRQs, iflib now binds those TX softirqs to CPUs using the above
mapping mechanism (that is, treats them as if they were TX IRQs).
Previously, such bindings were left up to the grouptaskqueue code and
thus fell outside of the iflib CPU mapping strategy.
Reviewed by: kbowling
Tested by: olivier, pkelsey
MFC after: 3 weeks
Differential Revision: https://reviews.freebsd.org/D24094
That is, provide wrappers around the atomic_testandclear and
atomic_testandset primitives.
Submitted by: jeff
Reviewed by: cem, kib, markj
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D22702
An upcoming patch to use the bitset macros for tracking vm page
dump information could conceivably need more than INT_MAX bits.
Expand the bit type to long so that the extra range is available
on 64-bit platforms where it would most likely be needed.
CPUSET_COUNT and DOMAINSET_COUNT are also modified to remain of
type `int`.
Reviewed by: kib, markj
Approved by: scottl (implicit)
MFC after: 1 week
Sponsored by: Ampere Computing, Inc.
Differential Revision: https://reviews.freebsd.org/D26190
s/BIT_NAND/BIT_ANDNOT/, and for CPU and DOMAINSET too. The actual
implementation is "and not" (or "but not"), i.e. A but not B.
Fortunately this does appear to be what all existing callers want.
Don't supply a NAND (not (A and B)) operation at this time.
Discussed with: jeff
Reviewed by: cem
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D22791
We have a couple optimizations for when the bitset is known to be just
one word. But with dynamically sized bitsets, it was actually more work
to determine the size than just to do the necessary computation. Now,
only use the optimization when the size is known to be constant.
Reviewed by: markj
Discussed with: jeff
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D22639
The _REENT_GLOBAL_STDIO_STREAMS was introduced by commit
668a4c8722 in 2017. Since then it was enabled by
default for RTEMS. Recently, the option was enabled for Cygwin which
previously used an alternative implementation to use global stdio streams.
In Newlib, the stdio streams are defined to thread-specific pointers
_reent::_stdin, _reent::_stdout and _reent::_stderr. If the option is disabled
(the default for most systems), then these pointers are initialized to
thread-specific FILE objects which use file descriptors 0, 1, and 2,
respectively. There are at least three problems with this:
(1) The thread-specific FILE objects are closed by _reclaim_reent(). This
leads to problems with language run-time libraries that provide wrappers to
the C/POSIX stdio streams (for example C++ and Ada), since they use the
thread-specific FILE objects of the initialization thread. In case the
initialization thread is deleted, then they use freed memory.
(2) Since thread-specific FILE objects are used with a common output device via
file descriptors 0, 1 and 2, the locking at FILE object level cannot ensure
atomicity of the output, e.g. a call to printf().
(3) There are resource managment issues, see:
https://sourceware.org/pipermail/newlib/2022/019558.htmlhttps://bugs.linaro.org/show_bug.cgi?id=5841
This patch enables the _REENT_GLOBAL_STDIO_STREAMS behaviour for all Newlib
configurations and removes the option. This removes a couple of #ifdef blocks.
For _REENT_GLOBAL_STDIO_STREAMS, lock/unlock all FILE objects. In the
repository, this function is only used by Cygwin during process forks. Since
Cygwin enabled _REENT_GLOBAL_STDIO_STREAMS recently, without this fix no FILE
object at all was locked.
Do not initialize __sglue with the FILE objects of _GLOBAL_REENT to avoid a
double use in the !_REENT_SMALL and !_REENT_GLOBAL_STDIO_STREAMS configurations
which didn't use a thread-specific reentrancy structure.
Remove "32" or "64" from each of the following names: acl32,
aclcheck32, aclfrommode32, aclfrompbits32, aclfromtext32, aclsort32,
acltomode32, acltopbits32, acltotext32, facl32, fchown32, fcntl64,
fstat64, _fstat64, _fstat64_r, ftruncate64, getgid32, getgrent32,
getgrgid32, getgrnam32, getgroups32, getpwuid32, getpwuid_r32,
getuid32, getuid32, initgroups32, lseek64, lstat64, mknod32, mmap64,
setegid32, seteuid32, setgid32, setgroups32, setregid32, setreuid32,
setuid32, stat64, _stat64_r, truncate64.
Remove prototypes and macro definitions of these names.
Remove "#ifndef __INSIDE_CYGWIN__" from some headers so that the new
names will be available when compiling Cygwin.
Remove aliases that are no longer needed.
Include <unistd.h> in fhandler_clipboard.cc for the declarations of
geteuid and getegid.
The __sFILE::_lock member is present if __SINGLE_THREAD__ is not defined. In
this case, it is initialized in __sfp(). It is a bug to do it sometimes also
in std().
The compiler warns the double parentheses are unnecessary in some
target, and cause fail cases when doing some testcases in regression.
gcc/testsuite/g++.dg/warn/Wstringop-overflow-6.C
Remove the unnecessary parentheses will fix it. See more details in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85775
Same like in commit 0542583129,
Author: Maxim Blinov <maxim.blinov@embecosm.com>
Date: Thu Jul 22 22:41:42 2021 +0100
Remove unneccesary parenthesis around declarator
Thanks for Sebastian Huber's remind!
For the exit processing only members of _GLOBAL_REENT were used by default. If
the _REENT_GLOBAL_ATEXIT option was enabled, then the data structures were
provided through dedicated global objects. Make this option the default.
Remove the option. Rename struct _reent members _atexit and _atexit0 to
_reserved_6 and _reserved_7, respectively. Provide them only if
_REENT_BACKWARD_BINARY_COMPAT is defined.
Rename struct _reent::_new::_unused members _nextf and _nmalloc to _reserved_3
and _reserved_4, respectively. Rename struct _reent::_new member _unused to
_reserved_5. Provide them only if _REENT_BACKWARD_BINARY_COMPAT is defined.
Remove unused _N_LISTS define.
Add the --enable-newlib-reent-binary-compat configure option. This option is
disabled by default. If enabled, then unused members in struct _reent are
preserved to maintain the structure layout.
Added a new global __sglue object for all configurations.
Decouples the global file object list from the _GLOBAL_REENT
structure by using this new object instead of the __sglue member
of _GLOBAL_REENT in __sfp() and _fwalk_sglue().
Replaced _fwalk_reent() with _fwalk_sglue(). The change adds an
extra __sglue object as a parameter, which will allow the passing
of a global __sglue object separate from the __sglue member of
struct _reent. The global __sglue object will be added in a
follow-on patch.
Define the configuration-dependent constant CLEANUP_FILE for use in
cleanup_stdio(). This will reduce duplicate code during the addition
of a dedicated stdio atexit handler in a follow-on patch.
Moved last remaining __sglue initializations from __sinit() to
__sfp(). The move better encapsulates access to __sglue and
facilitates its decoupling from struct _reent in a follow-on patch.
Remove __sinit_lock_acquire() and __sinit_lock_release(). Replace these with
__sfp_lock_acquire() and __sfp_lock_release(), respectively. This eliminates a
potential deadlock issue between __sinit() and __sfp(). Remove now unused
__sinit_recursive_mutex and __lock___sinit_recursive_mutex.
Added _REENT_INIT_SGLUE and _REENT_INIT_SGLUE_ZEROED macros
to initialize __sglue member of struct _reent. This allows
further simplification of __sinit() and facilitates the removal
of __sglue as a member of struct _reent for certain configurations
in a follow-on patch.
Removed duplicate sglue initializations from __sinit(). These
are already initialized in the _REENT_INIT macro in sys/reent.h.
This simplification enables the reduction of _GLOBAL_REENT
dependency in a follow-on patch.
Removed duplicate stdio initializations from __sinit(). These
are already initialized in the _REENT_INIT macro in sys/reent.h.
This simplification enables the reduction of _GLOBAL_REENT
dependency in a follow-on patch.
../newlib/libc/libc.xml:22242: element refentry: validity error : ID iconv already defined
<refentry id="iconv">
Use a separate namespace for chaper ids, to avoid collision between the
ids for the chapter and function 'iconv', now that iconv documentation
is generated unconditionally.
Unless make is invoked with V=1, have xmlto pass the parameter
'man.output.quietly=1' to xsltproc to suppress "Note: Writing foo.N"
output from the manpages stylesheet.
(This doesn't quite do what it says: The output is not silenced if V has
any value, including 0. You could consider that either a bug or a
feature.)
Simplify rules for creating docbook XML used to create manpages:
Updating the output using move-if-change and then unconditionally
touching the .stamp file doesn't make much sense.
The ndbm.c build broke with:
Commit 357d7fcc6
In <stdio.h> provide only necessary types
The above commit exposed a latent missing-header bug:
newlib/newlib/libc/include/ndbm.h:83:38: error: unknown type name ‘mode_t’
Signed-off-by: Dimitar Dimitrov <dimitar@dinux.eu>
The nano malloc build broke with:
Commit 357d7fcc6
In <stdio.h> provide only necessary types
The above commit exposed a latent missing-header bug:
newlib/libc/stdlib/nano-mallocr.c:220:33: error: ‘uintptr_t’ undeclared (first use in this function)
Fix by including <stdint.h>.
Signed-off-by: Dimitar Dimitrov <dimitar@dinux.eu>
Remove the pointer indirection through the read-only _global_impure_ptr and
directly use a externally visible _impure_data object of type struct _reent.
This enables the static initialization of global data structures in a follow up
patch. In addition, we get rid of a machine-specific file.
By including sys/_stdint.h, all types from stdint.h are
exposed even if stdint.h isn't pulled in explicitely. Include
<machine/_default_types.h instead. Fix up newlib and Cygwin
files which rely on stdint.h types, too.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
- from Brian Inglis <Brian.Inglis@systematicsw.ab.ca>
- add support to _tzset_unlocked_r() to support quoting std and dst
names with angle brackets <> as per Posix
- modify documentation of tzset.c
This fixes a bug introduced in a previous patch (Commit 44b60f0c:
Make __sdidinit unused). Removed intitialization of __cleanup from
__smakebuf_r(). All callers of __smakebuf_r() call __sinit()
through the_CHECK_INIT macro, thus __cleanup is already
initialized. This fix also allows _cleanup_r() to be made static.
Changed its name to cleanup_stdio() and removed its declaration
from local.h.
Rename __sfmoreglue() in sfmoreglue() and make it static. This function is
only used by __sfp() in the same translation unit.
Remove use of register keyword.
[PATCH] newlib: Only call _fputwc_r if ELIX_LEVEL >= 4
(nano-)vfprintf.c is enabled for ELIX_LEVEL >= 1. When _WIDE_ORIENT
is set, its __sprint_r / __sfputs_r functions unconditionally called
_fputwc_r which is only in ELEX_LEVEL >= 4. With this commit,
the _WIDE support in (nano-)vfprintf.c is disabled for ELEX_LEVEL < 4.
This code has not been updated since 2016, and it looks like it has
rotted quite a bit since. It does not build against the current set
of phoenix sources -- I had to hack both the kernel headers and the
newlib headers up to get it to build, and I still have no idea if it
actually links or runs. It seems like the project itself has moved
away from newlib and to its own C library:
https://phoenix-rtos.com/documentation/libc/README.md
So since there's no interest from the phoenix folks to maintain this,
and it has a significant amount of non-standard code that we try to
keep up-to-date (without actually testing it), just punt it all.
The recent makefile reorganization broke the amdgcn port by creating
duplicate __malloc_lock symbols. This patch fixes the problem by renaming
the malloc_support.c file to mlock.c, thus overriding the default symbol
properly. Actually, I'm not sure how this ever worked?
I've had this lying around for probably a year or two at this point.
It just changes all the instance of "errno" from a common symbol to an
extern. I can't offhand recall where the actual definition is, but it
certainly exists in the generic code.
This was disabled as part of the migration away from the cygnus option
as that implied no-dependencies.
We currently have 1-to-1 updates enabled -- if you touch a .c file,
the corresponding .o file will be rebuilt. But if you touch a header
file, none of the files using that get rebuilt.
Integrate the old libm/test/ subdir into the main build. It hasn't
been used in a long time causing the code to rot a bit. I've fixed
some of those, but it still fails for many ports, so it's disabled
by default. People who want to take a closer look can run:
$ make libm/test/test
To help prevent people from missing running this script, integrate it
into the build via maintainer mode.
Also fix the inverted exit status to make this work correctly -- for
some reason, it exited 1 when it worked, and 0 when it failed.
Replace all of the individual autotool steps with a single autoreconf.
This simplifies the documentation greatly, and in the current system,
only takes ~10 seconds to regenerate everything.
Update the developer documentation to cover all the major components
of the current build system. Hopefully this is a fairly complete road
map to everything. I tried to include everything that I wish I knew
when I started hacking on this :P.
Convert all the libc/ subdir makes into the top-level Makefile. This
allows us to build all of libc from the top Makefile without using any
recursive make calls. This is faster and avoids the funky lib.a logic
where we unpack subdir archives to repack into a single libc.a. The
machine override logic is maintained though by way of Makefile include
ordering, and source file accumulation in libc_a_SOURCES.
There's a few dummy.c files that are no longer necessary since we aren't
doing the lib.a accumulating, so punt them.
The winsup code has been pulling the internal newlib ssp library out,
but that doesn't exist anymore, so change that to pull the objects.
Rather than define per-object rules in the Makefile, have small files
that define & include the right content. This simplifies the build
rules, and makes understanding the source a little easier (imo) as it
makes all the subdirs behave the same: you have 1 source file and it
produces 1 object. It's also about the same amount of boiler plate,
without having to define custom build rules that can fall out of sync.
We also realign the free & pvalloc definitions: common code puts these
in malloc.o & valloc.o respectively, not in free.o & pvalloc.o objects.
This will also be important as we merge the libc.a build into the top
dir since it relies on a single flat list of objects for overrides.
The mallopt symbol is defined in tiny-malloc.c, not mallocr.c, but
the Makefile in here tries to compile it out of the latter. This
leads to mallopt never being defined.
The build also creates mallinfo.o & mallopt.o & mallstats.o objects
to override common ones, but the common dir doesn't use these names.
Instead, it places these all in mstats.o.
So move the build define logic to a dedicated file and compile it
directly to make things a bit simpler while fixing the missing func
and aligning objects with the cmomon code.
Rather than define per-object rules in the Makefile, have small files
that define & include the right content. This simplifies the build
rules, and makes understanding the source a little easier (imo) as it
makes all the subdirs behave the same: you have 1 source file and it
produces 1 object. It's also about the same amount of boiler plate,
without having to define custom build rules that can fall out of sync.
This will also be important as we merge the libc.a build into the top
dir since it relies on a single flat list of objects for overrides.
Also take the opportunity to clean up the unnecessary header deps in
here. Automake provides dependency generation for free now.
Some awk implementations such as old versions of mawk do not support the
length() function. Use the return value of the POSIX split() function instead.
This file is a little confusing: it provides all of the mallocr logic,
but is compiled multiple times to produce a unique symbol each time.
For example, building mallocr.c with -DDEFINE_FREER produces freer.o
that only defines _free_r(). This is fine for most symbols, but it's
a little confusing when defining mallocr itself -- we produce a file
with the same symbol name, but we still need -DDEFINE_MALLOCR. In
order to move the logic from the build rules to source files, using
mallocr.c both as a multiplexer and for defining a single symbol is a
bit tricky. It's possible (if we add a lot of redundant preprocessor
checks to mallocr.c, or we add complicated build flags just for this
one files), but it's easier if we simply rename this to a dedicated
file. So let's do that.
We do this as a dedicated commit because the next one will create a
new mallocr.c file and git's automatic diff algorithms can handle
trivial renames, but it can't handle renames+creates in the same
commit.
Simplify the build system logic a bit by moving the mallocr.c ->
nano-mallocr.c redirection from the Makefile to the source files.
This allows for consistent object name usage regardless of the
configuration options used in case a machine dir wants to define
its own override.
Since we already set up _LIBC to indicate source files are building
for newlib, we don't need this malloc-specific symbol. Convert it
over to simplify the build a bit.
The current targ-include setup runs `cp` every header file it installs,
in serial. This can be a little noticeable on systems, so cleanup the
logic to rely on cp's ability to copy multiple files to a directory in
a single call.
We still need a check for empty directories with no headers (i.e. the
glob doesn't match anything), so add a helper variable to contain that
logic to reduce the boiler plate a little.
Rather than define per-object rules in the Makefile, have small files
that define & include the right content. This simplifies the build
rules, and makes understanding the source a little easier (imo) as it
makes all the subdirs behave the same: you have 1 source file and it
produces 1 object. It's also about the same amount of boiler plate,
without having to define custom build rules that can fall out of sync.
Some of these rules were already unnecessary as they were compiling a
single source file into the same named object w/out custom flags, and
Automake handles that for us completely.
This will also be important as we merge the libc.a build into the top
dir since it relies on a single flat list of objects for overrides.
Also take the opportunity to clean up the unnecessary header deps in
here. Automake provides dependency generation for free now.
Since POSIX cp requires copying a file to a directory without having
to specify the name explicitly, rely on that to avoid calling basename
on every source file.
We can also drop the stub `true` call if the -f test failed. The use
of `if` already takes care of that in POSIX shell.
This kills off the last configure script under libc/ and folds it
into the top newlib configure script. The a lot of the logic was
already in the top configure script, so move what's left into a
libc/acinclude.m4 file.
I split libg.a out into a sep target from libc.a for the main dir in
commit f2b053f49e ("newlib: separate out
libg from libc"), but missed the multilib dirs. That leads to an
uncommon parallel build failure:
- libc.a rule runs & finishes
- $(BUILD_MULTISUBDIR)/libc.a rule runs
-> failure due to libg.a not yet existing
- libg.a rule runs & finishes
Split the multilib libg rule out from libc too so it can depend on the
main libg directly and avoid this race.
Remove dependency on __sdidinit member of struct _reent to check
object initialization. Like __sdidinit, the __cleanup member of
struct _reent is initialized in the __sinit() function. Checking
initialization against __cleanup serves the same purpose and will
reduce overhead in the __sfp() function in a follow up patch.
The crt0.o was handled in a subdir-by-subdir basis: it would be compiled
in one (e.g. libc/sys/$arch/), then copied up one level (libc/sys/), then
copied up another (libc/) before finally being copied & installed in the
top newlib dir. The libc/sys/ copy was cleaned up, and then the top dir
was changed to copy it directly out of the libc/sys/$arch/ dir. But the
libc/sys/ copy to libc/ was left behind. Clean that up now too.
These headers aren't installed, so use "" includes instead of <> so
we don't search system header paths. This matches the style used
elsewhere in the tree for these local headers, and makes it work
w/out explicit -I flags (as needed with non-recursive make).
Convert all the libm/ subdir makes into the top-level Makefile. This
allows us to build all of libm from the top Makefile without using any
recursive make calls. This is faster and avoids the funky lib.a logic
where we unpack subdir archives to repack into a single libm.a. The
machine override logic is maintained though by way of Makefile include
ordering, and source file accumulation in libm_a_SOURCES.
One thing to note is that this will require GNU Make because of:
libm_a_CFLAGS = ... $(libm_a_CFLAGS_$(subst /,_,$(@D)))
This was the only way I could find to supporting per-dir compiler
settings, and I couldn't find a POSIX compatible way of transforming
the variable content. I don't think this is a big deal as other
Makefiles in the tree are using GNU Make-specific syntax, but I call
this out as it's the only one so far in the new automake code that
I've been writing.
Automake doesn't provide precise control over the output object names
(by design). This is fine by default as we get consistent names in all
the subdirs: libm_a-<source>.o. But this relies on using the same set
of compiler flags for all objects. We currently compile libm/common/
with different optimizations than the rest.
If we want to compile objects differently, we can create an intermediate
archive with the subset of objects with unique flags, and then add those
objects to the main archive. But Automake will use a different prefix
for the objects, and thus we can't rely on ordering to override.
But if we leverage $@, we can turn Automake's CFLAGS into a multiplex
on a per-dir (and even per-file if we wanted) basis. Unfortunately,
since $@ contains /, Automake complains it's an invalid name. While
GNU Make supports this, it's a POSIX extension, so Automake flags it.
Using $(subst) avoids the Automake warning to get a POSIX compliant
name, albeit with a GNU Make extension.
Make this a separate target from libc so that we can migrate libc over
to automake more easily. Having it integrated into the libc target is
difficult to handle when using automake rules which expect a one-to-one
mapping between names & inputs.
When migrating the manual to the top-level, the include order was
sorted by name of the subdir. But this changed the chapter order
of the manual in the process. Change the sorting back to match
existing chapters and update the comments to explain.
Using xxx_LIBADD, xxx_DEPENDENCIES, and EXTRA_xxx_SOURCES is one way of
conditionally including files into a target. But it's meant more for the
cases where the variables added to LIBADD & DEPENDENCIES are constructed
via substitution (e.g. AC_SUBST) or other dynamic methods. With Automake
conditionals, then the much simpler form is to conditionally append to
the xxx_SOURCES variable and let Automake sort everything out.
Commit 8fa73a9f84 changed how fenv.c is
compiled wrt mips16 targets used the wrong variable to add fenv.o to
libm.a. Fix that thinko so it's included in the build again.
The top-level newlib dir already takes care of recursing into the
sys/xxx/include/ subdirs and installing any headers found, so the
rtems subdir doesn't need to do this itself.
Replace the custom build rules (which require copying & pasting from the
current Makefile) with small stub files. This allows us to drop the rules
entirely and let Automake provide everything.
These subdirs don't actually use anything from libm. The common dir
in particular only has 4 header files, and none are included here.
The xstormy16 code has a comment mentioning why this hack is here, but
it refers to code that was removed when its configure script was merged
up a level.
This is used in a bunch of places, but nowhere is it ever set, and
nowhere can I find any documentation, nor can I find any other project
using it. So delete the flags to simplify.
These targets don't actually cross-compile -- they try to pull some
objects out of the host's /lib/libc.a, /lib/libm.a, and /lib/crt0.o
directly and merge them into newlib's own libraries. This is hard
to keep working and impossible to test. Considering the vintage of
such targets, and gcc dropping them many many years ago, drop them
from newlib too. This will make cleaning up the build a lot easier.
The machine/{configure,Makefile} files exist only to fan out to the
specific machine/$arch/ subdir. We already have all that same info
in the phoenix/ dir itself, so by moving the recursive configure and
make calls into it, we can cut off this logic entirely and save the
overhead.
These were never added to the tree, and as we transition from autoconf
to automake, it really wants the latter subdirs to always exist. These
don't, so delete the logic.
The original cut for small arguments at |x|<2**-70 (copied from the
double version) produces that when computing nadj we get a subnormal
number for t*x and thus, the division of pi/subnormal will be INF and
the logarithm of it too, which is wrong as a result for lgammaf in this
range.
The proposed new limit seems to be safe and has been tested to
produce accurate results.
(Courtesy of Andreas Jung, ESA)
These look like they were just copied & pasted from common/Makefile.am.
The funcs in this dir are all stubs that don't actually call any math
or builtin functions, and a simple compile shows they produce identical
object code. So delete to simplify the build rules.
Correct the overflow limit in the variable o_threshold to be consistent
with the FLT_UWORD_LOG_MAX variable used by the internal implementation
of the expf algorithm itself.
The u_threshold variable has also been modified to be written in the
same format.
Note that this fix improves the situation but does not completely
correct the inconsistencies regarding the overflow and underflow limits
between the expf wrapper (wf_exp.c) and the expf algorithm itself
(ef_exp.c).
Currently these limits are different for the
_FLT_LARGEST_EXPONENT_IS_NORMAL and _FLT_NO_DENORMALS cases as well as
for the case where __OBSOLETE_MATH is not defined (only for the
underflow limit in this case).
This kills off the last configure script under libm/ and folds it
into the top newlib configure script. The vast majority of logic
was already in the top configure script, so move the little that
is left into a libm/acinclude.m4 file.
Make sure we depend on the right name of mkdoc all the time, and that
the rules that need it (e.g. .def files) depend on it.
Reported-by: Jon Turney <jon.turney@dronecode.org.uk>
This was only ever used for i?86-pc-linux-gnu targets, but that's been
broken for years, and has since been dropped. So clean this up too.
This also deletes the funky objectlist logic since it only existed for
the libtool libraries. Since it was the only thing left in the small
Makefile.shared file, we can punt that too.
This was only used by the i?86-pc-linux-gnu target which we've removed,
and even though it's using a "sys/linux/" dir to make it sound like it
only depends on the Linux kernel, it's actually tied to glibc APIs built
on top of Linux. Since the code relies on internal glibc APIs and has
been broken for some time, punt it all. If someone wants to bring it
back, they can try and actually keep the Linux-vs-glibc APIs separate.
This was added 20+ years ago. It seems to have very few (or no users)
as it only works on 32-bit x86 GNU/Linux (i.e. glibc) systems, and even
then only with old versions of glibc. It hasn't compiled in at least 5
years, but most likely been broken for more like 15 years -- it relies
on internal glibc APIs (like linuxthreads), and that code has changed
and been deleted significantly since.
This single target ends up dragging in a lot of non-trivial code that is
hard to keep working, and currently impossible to verify -- the libtool
and iconvdata and sys/linux/ code isn't used by anything else, but ends
up touching just about every build file in the tree. Punt the target so
we can start stripping out all these unique code paths.
This commit by itself just disables the target. We'll start deleting the
individual unused pieces in followups.
Now that we use AC_NO_EXECUTABLES, and we require a recent version of
autoconf, we don't need to define our own copies of these macros. So
switch to the standard AC_PROG_CC.
This logic was added to libc & libm to get it working again after some
reworks in the CPP handling, but now that that's settled, let's move
this to the common newlib configure logic. This will make it easier
to consolidate all the configure calls into the top-level newlib dir.
This does create a lot of noise in the generate scripts, but that's
because of the ordering of the calls, not because of correctness. We
will try to draw that back down in follow up commits as we modernize
the toolchain calls in here.
This code is a bit more convoluted than it needs to be. GPR_SOURCES
is never set to anything, and the automake checks use negative logic
to add the SP & DP source files to dedicated variables that are only
expanded once. Get rid of the unused variable, use normal boolean
logic, and collapse the source settings into a single variable.
This allows building the libc & libm pages in parallel, and drops
the duplication in the subdirs with the chew/chapter settings.
The unused rules in Makefile.shared are left in place to minimize
noise in the change.
This doesn't migrate all the docs, just the libc's manual (pdf/info).
This is to show the basic form of migrating the chew files.
For subdirs that didn't have any docs, I've stripped their settings
for clarity. If someone wanted to suddenly add docs, they can add
the corresponding Makefile.inc files easily.
THe stdio subdir is actually required by the documentation. The
stdio/def is handled dynamically, but libc.texi always expects it
to be included, and fails if it isn't. So making it required when
building docs is safe.
The xdr subdir is handled dynamically, but it doesn't include any
docs, so the dynamic logic isn't (currently) adding any value. So
making it required when building docs is safe.
That leaves: iconv, stdio64, posix, and signal subdirs. The chapters
have a little disclaimer saying they are system-dependent, but even
then, imo having stable manuals regardless of the target is preferable,
and we can add more disclaimer language to these chapters if we want.
This doesn't touch the man page codepaths, just the info/pdf.
Let automake manage whether the objects are included in lib.a. This
fixes failures after to commit 71086e8b2d
("newlib: delete (most) redundant lib_a_CCASFLAGS=$(AM_CCASFLAGS)") due
to automake generating different set of implicit rules, and the code in
here assuming the names of the generated objects.
We've been using both libc_cv_ and newlib_cv_ for our cache vars.
Let's consolidate on newlib_cv_ to avoid conflicts with glibc which
is already using the libc_cv_ prefix.
This isn't strictly necessary, but it makes for much clearer logs as
to what the target is doing, and provides cache vars for anyone who
wants to force the test a different way, and it lets the build cache
its own results when rerunning config.status.
Since commit dcbff9eea7 ("newlib: merge
iconvdata into top-level Makefile"), there is no configure script in
the iconvdata/ subdir, so this call will just issue a warning and not
do anything useful. Punt it.
Restore the call to AC_NO_EXECUTABLES -- I naively assumed in commit
2e9aa5f56c ("newlib: update preprocessor
configure checks") that checking for a preprocessor would not involve
linking code. Unfortunately, autoconf will implicitly check that the
compiler "works" before allowing it to be used, and that involves a
link test, and that fails because newlib provides the C library which
is needed to pass a link test.
There is some code in NEWLIB_CONFIGURE specifically to help mitigate
these, but it's not kicking in here for some reason, so let's just add
the AC_NO_EXECUTABLES call back until we can unwind that custom logic.
Additionally, we have to call AC_PROG_CPP explicitly. This was being
invoked later on, but only in the use_libtool=yes codepath, and that
is almost never enabled.
This code snippet assumed it was only ever run in the top configure
script where srcdir would point to newlib/ which is parallel to the
winsup/ tree. This is incorrect for all of the subdir configure
scripts leading to bad -I flags in $(CC). Switch it over to the
new abs_newlib_basedir which should work in all subdirs.
When we had configure scripts in subdirs, the newlib_basedir value
was computed relative to that, and it'd be the same when used in the
Makefile in the same dir. With many subdir configure scripts removed,
the top-level configure & Makefile can't use the same relative path.
So switch the subdir Makefiles over to abs_newlib_basedir when they
use -I to find source headers.
Do this for all subdirs, even ones with configure scripts and where
newlib_basedir works. This makes the code consistent, and avoids
surprises if the configure script is ever removed in the future as
part of merging to the higher level.
Some of the subdirs were using -I$(newlib_basedir)/../newlib/ for
some reason. Collapse those too since newlib_basedir points to the
newlib source tree already.
When using the top-level configure script but subdir Makefiles, the
newlib_basedir value gets a bit out of sync: it's relative to where
configure lives, not where the Makefile lives. Move the abs setting
from the top-level configure script into acinclude.m4 so we can rely
on it being available everywhere. Although this commit doesn't use
it anywhere, just lays the groundwork.
Commit dd23de27c8 ("newlib: libc: install
CRT0 straight out of subdir") got rid of the libc/sys/ intermediate for
copying the file up, but the top-level newlib/ dir was still expecting
a libc/crt0.o to exist so it could install. Update that to also look
for the crt0 file directly under libc/ like we already do for crt1.
The work to merge libc/machine/ up a dir lost the stub doc targets.
So when libc/ recursed into machine/, it would stop going deeper as
the doc rules were empty. But now that libc/ goes directly into the
libc/machine/$arch/ and those have never had doc stubs, the build
fails. Add a quick hack to the top dir to ignore all machine/$arch/
dirs when generating docs. A follow up series will delete all of
this code as it merges all the doc rules into the top newlib dir.
We use the common config-ml.in for configure, so switch the makefile
over to the common multilib.am. It's almost exactly the same code,
but there are two differences:
* Common code hooks install-exec-local for install-multi, but newlib
doesn't currently install any executables, so that doesn't fire.
Newlib already has install-data-local that inlined install-multi,
so switch that to the common install-multi.
* Common code doesn't provide a check-multi at all. Keep ours for
now. Some day common code might get it. Or not. Who knows.
The machine configure scripts are all effectively stub scripts that
pass the higher level options to its own makefile. There were only
three doing custom tests. The rest were all effectively the same as
the libc/ configure script.
So instead of recursively running configure in all of these subdirs,
generate their makefiles from the top-level configure. For the few
unique ones, deploy a pattern of including subdir logic via m4:
m4_include([machine/nds32/acinclude.m4])
Some of the generated machine makefiles have a bunch of extra stuff
added to them, but that's because they were inconsistent in their
configure libtool calls. The top-level has it, so it exports some
new vars to the ones that weren't already.
The sys configure scripts are almost all effectively stub scripts that
pass the higher level options to its own makefile. The phoenix & linux
ones are a bit more complicated with nested subdirs, so those have been
left alone for now. Plus, I don't really have a way of testing them.
There's no need to have a sys/ subdir just to copy the sys/$arch/crt0.o
up to sys/crt0.o, and then have libc/ copy sys/crt0.o up again. Just
have libc/ refer to sys/$arch/crt0.o directly and drop the intermediate
makefile entirely.
The sys/{configure,Makefile} files exist to fan out to the specific
sys/$arch/ subdir, and to possibly generate a crt0. We already have
all that same info in the libc/ dir itself, so by moving the recursive
configure and make calls into it, we can cut off some of this logic
entirely and save the overhead.
For arches that don't have a sys subdir, it means they can skip the
logic entirely.
The sys subdir itself is kept for the crt0 logic, for now. We'll try
and clean that up next.
The machine/{configure,Makefile} files exist only to fan out to the
specific machine/$arch/ subdir. We already have all that same info
in the libc/ dir itself, so by moving the recursive configure and
make calls into it, we can cut off this logic entirely and save the
overhead.
For arches that don't have a machine subdir, it means they can skip
the logic entirely. Although there's prob not too many of those.
The machine configure scripts are all effectively stub scripts that
pass the higher level options to its own makefile. The only one doing
any custom tests was nds32. The rest were all effectively the same as
the libm/ configure script.
So instead of recursively running configure in all of these subdirs,
generate their makefiles from the top-level configure. For nds32,
deploy a pattern of including subdir logic via m4:
m4_include([machine/nds32/acinclude.m4])
Even its set of checks are very small -- it does 2 preprocessor tests
and sets up 2 makefile conditionals.
Some of the generated machine makefiles have a bunch of extra stuff
added to them, but that's because they were inconsistent in their
configure libtool calls. The top-level has it, so it exports some
new vars to the ones that weren't already.
The machine/{configure,Makefile} files exist only to fan out to the
specific machine/$arch/ subdir. We already have all that same info
in the libm/ dir itself, so by moving the recursive configure and
make calls into it, we can cut off this logic entirely and save the
overhead.
For arches that don't have a machine subdir, it means they can skip
the logic entirely.
This makes the makefile logic a bit cleaner so we don't have two
files maintaining lists of sources & objects. Since the logic is
tied to cpu capabilities, past those boolean settings down from
the configure logic to the makefile logic.
This will also make it easier to throw away the configure script
in a follow up commit and just keep the makefile.
The nds32 & spu dirs are using compile tests to look for some
preprocessor defines, but we don't need to compile the code,
just preprocess it. So switch to AC_PREPROC_IFELSE.
The sh dir is using a preprocessor test via grep, but let's
switch it to AC_PREPROC_IFELSE too to be consistent.
This should allow us to drop the uncommon AC_NO_EXECUTABLES call.
This was added decades ago, but the commit message lacks any
explanation, and it was unused when it was merged. It's still
unused today. So punt it all.
Generating these files is very cheap, so let's just do it all the time.
This makes the build logic simpler, and keeps errors for slipping in in
codepaths that are not well tested. Creating these files doesn't mean
they'll be included in the manual implicitly.
For example, some of the nano stdio files break documentation because
they don't have any chew directives in them. But no one noticed since
that code path is rarely enabled. So drop the _i and _float def files.
This is already handled by autotools for us automatically. You can
tell as the generated output is exactly the same other than deleting
a few blank lines.
It's unclear why this was added originally, but assuming it was needed
20 years ago, it shouldn't be explicitly required nowadays. Current
versions of autotools already take care of exporting LDFLAGS to the
Makefile as needed (things are actually getting linked). That's why
the configure diffs show LDFLAGS still here, but shifted to a diff
place in the output list. A few dirs stop exporting LDFLAGS, but
that's because they don't do any linking, only compiling, so it's
correct.
As for the use of $ldflags instead of the standard $LDFLAGS, I can't
really explain that at all. Just use the right name so users don't
have to dig into why their setting isn't respected, and then use a
non-standard name instead. Adjust the testsuite to match.
The current newlib multilib logic is almost exactly the same as the
config/multi.m4, and the differences should be minor, so switch over
to that to delete custom logic on ourside.
Now that we require a recent version of autoconf, we can rely on this
macro working. This change was already made to libm, but these other
dirs were missed as I didn't notice it being duplicated in 3 places.
Now that newlib.hin has been brought up to date and all of its defines
are produced by configure, we can switch it to using autoheader without
manual editing. This relies on a few pieces:
* Moving the header & footer into configure.ac via AH_TOP & AH_BOTTOM.
* Running a post-process step on newlib.h to delete all the defines we
didn't export ourselves. Basically, anything without a _ prefix.
This will leave behind some spurious comments in newlib.h related to
the defines we filtered out, but should be harmless, so it's probably
not worth the effort to construct a more complicated sed expression to
also strip those out.
Alan Somers
Brooks Davis
Pedro F. Giffuni
Sebastian Huber
Warner Losh
Mark Johnston
Alex Richardson
Ken Brown
Stefan Eßer
Konstantin Belousov
Patrick Kelsey
Ryan Libby
D Scott Phillips
Jeff Roberson
jdoubleu
Jeff Johnston
Jia-Wei Chen
Matt Joyce
Jon Turney
Dimitar Dimitrov
Corinna Vinschen
Andrea Corallo
Mike Frysinger
Andrew Stubbs
Jeff Law
Andoni Arregi
Cyril Yared