OSDN Git Service

uclinux-h8/uclibc-ng.git
20 years agoAllow building on 64 bit archs. Hopefully the last cvs activity this
Manuel Novoa III [Thu, 5 Feb 2004 02:38:27 +0000 (02:38 -0000)]
Allow building on 64 bit archs.  Hopefully the last cvs activity this
stuff will see other than "delete".

20 years agoUse 'mmap.c' instead of '_mmap.c'
Eric Andersen [Thu, 5 Feb 2004 02:17:02 +0000 (02:17 -0000)]
Use 'mmap.c' instead of '_mmap.c'

20 years agoFixup a badly merged update
Eric Andersen [Thu, 5 Feb 2004 01:54:33 +0000 (01:54 -0000)]
Fixup a badly merged update
 -Erik

20 years agoUse a static const char array for static strings
Eric Andersen [Wed, 4 Feb 2004 10:24:12 +0000 (10:24 -0000)]
Use a static const char array for static strings

20 years agoToday's Toni's birthday. Seems an appropriate day to add this.
Manuel Novoa III [Tue, 3 Feb 2004 17:33:50 +0000 (17:33 -0000)]
Today's Toni's birthday.  Seems an appropriate day to add this.

20 years agoUse the correct configuration test define.
Manuel Novoa III [Mon, 2 Feb 2004 22:27:22 +0000 (22:27 -0000)]
Use the correct configuration test define.

20 years agoThe code for checking the rpath was not finding the rpath
Eric Andersen [Sun, 1 Feb 2004 11:10:32 +0000 (11:10 -0000)]
The code for checking the rpath was not finding the rpath
value correctly, since it forgot to lookup the correct value
in the string table.

20 years agoJakub Bogusz writes:
Eric Andersen [Sun, 1 Feb 2004 08:30:55 +0000 (08:30 -0000)]
Jakub Bogusz writes:

Hello,

the attached patch fixes ldso compilation on sparc-linux (at least with
sparc32 userland). It was made against uClibc 0.9.21, but it's still OK
with 0.9.26.

There were three problems:
- missing __NR_getpid definition
- conflicting prototypes in elfinterp.c
- syntax error (missing semicolon) in ldso.c

With this patch shared libraries and ldso can be built on sparc.

It contains also ldso.c syntax fix for m68k - there was the same
mistake two lines earlier; but I didn't test build on m68k.

20 years agoRevert to scope as used by original code for now, just in case.
Eric Andersen [Sat, 31 Jan 2004 03:15:07 +0000 (03:15 -0000)]
Revert to scope as used by original code for now, just in case.
I still don't have a good handle on why and when the different
scope values should be used.

20 years agoPatch from Paul Mundt, lethal at linux-sh dot org,
Eric Andersen [Sat, 31 Jan 2004 03:02:39 +0000 (03:02 -0000)]
Patch from Paul Mundt, lethal at linux-sh dot org,
add sh/sh64: ABORT_INSTRUCTION definitions

20 years agoAndrew May writes:
Eric Andersen [Fri, 30 Jan 2004 22:10:39 +0000 (22:10 -0000)]
Andrew May writes:

Here are some simple fixes for things that broke for PPC with
the recent syscall cleanup. I am not sure they are correct but
they seem pretty trivial.

20 years agoAdd missing abort instruction for powerpc
Eric Andersen [Fri, 30 Jan 2004 21:13:24 +0000 (21:13 -0000)]
Add missing abort instruction for powerpc

20 years agoPaul Mundt, lethal at linux-sh dot org writes:
Eric Andersen [Fri, 30 Jan 2004 21:11:18 +0000 (21:11 -0000)]
Paul Mundt, lethal at linux-sh dot org writes:

Boredom got the better of me, here's strlen() for sh64..

20 years agoJoakim Tjernlund writes:
Eric Andersen [Fri, 30 Jan 2004 20:34:03 +0000 (20:34 -0000)]
Joakim Tjernlund writes:

> Not there yet, but the interfaces are much closer now...
> Heading to bed though, since its 5:30am.  :-)

This works, but I discovered something "funny". For all
relocs but COPY you can use scope instead of scope->dyn->symbol_scope
and it is much faster to do that. Search for "Funny" in the patch
to see what I mean. Probably I probably broke something, please let me
know if it works for you too.

I think I am done with ldso now. Has been fun and I hope to
actually use uClibc/busybox soon in a project.

Ohh, there is one thing left, double sized PLT entries, but maybe these
aren't needed in uClibc?

 Sweet dreams
              Jocke

20 years agoModify interfaces for _dl_parse_relocation_information()
Eric Andersen [Thu, 29 Jan 2004 12:18:58 +0000 (12:18 -0000)]
Modify interfaces for _dl_parse_relocation_information()
_dl_parse_lazy_relocation_information() and _dl_parse_copy_information()
so they are all consistant, allowing for future consolidation.

Trim some trailing whitespace as well.

20 years agoScrub up use of ELF_USES_RELOCA and eliminte some unsightly ifdefs
Eric Andersen [Thu, 29 Jan 2004 11:42:46 +0000 (11:42 -0000)]
Scrub up use of ELF_USES_RELOCA and eliminte some unsightly ifdefs

20 years agoJoakim Tjernlund writes:
Eric Andersen [Thu, 29 Jan 2004 10:48:05 +0000 (10:48 -0000)]
Joakim Tjernlund writes:

Here is the cleaned up laze reloc patch.
Summary:
 - Minor cleanup.
 - disable the "if (finaladdr <= 0x01fffffc || finaladdr >= 0xfe000000)"
   test since it almost never triggered.
 - Optimized the lazy relocs handling.

Would be great if you could commit ASAP.

 Jocke

20 years agoEliminate separate passes for _dl_copy_fixups() and _dl_fixup(), and
Eric Andersen [Thu, 29 Jan 2004 10:44:50 +0000 (10:44 -0000)]
Eliminate separate passes for _dl_copy_fixups() and _dl_fixup(), and
do both operations in a single pass.

20 years agoFixup silly typo
Eric Andersen [Thu, 29 Jan 2004 08:59:00 +0000 (08:59 -0000)]
Fixup silly typo

20 years agoPatch from Joakim Tjernlund:
Eric Andersen [Thu, 29 Jan 2004 08:20:11 +0000 (08:20 -0000)]
Patch from Joakim Tjernlund:

Removed R_PPC_NONE and R_PPC_REL24 as these does not seem to be used.
Corrected R_PPC_ADDR32 and R_PPC_JMP_SLOT

20 years agoPatch from Joakim Tjernlund to get rid of an annoying warning.
Eric Andersen [Tue, 27 Jan 2004 07:37:12 +0000 (07:37 -0000)]
Patch from Joakim Tjernlund to get rid of an annoying warning.

20 years agoJoakim Tjernlund writes:
Eric Andersen [Tue, 27 Jan 2004 07:36:19 +0000 (07:36 -0000)]
Joakim Tjernlund writes:

Hi Erik

I have had some fun trying to optimize memcpy, memset and memmove for
PPC. There are only boot tested, but  I don't expect any problems :)

Read the comments in powerpc/string.c for more info.
Patch is relative to libc/string

 Jocke

20 years agoOleksandr G. Zhadan writes:
Eric Andersen [Sat, 24 Jan 2004 08:40:27 +0000 (08:40 -0000)]
Oleksandr G. Zhadan writes:

Hi Eric,

The attached diff file includes BUS_ISA fix for kernel since 2.4.23/

Kind regards,

Oleks

20 years agoImre Sunyi writes:
Eric Andersen [Sat, 24 Jan 2004 08:33:39 +0000 (08:33 -0000)]
Imre Sunyi writes:

Hi Erik

I have corrected a bug in uClibc/libc/inet/resolv.c in function
__dns_lookup(). Have attaced a txt file with my diffs regarding to
uClibc 0.9.26.

If two nameservers are included in /etc/resolv.conf and the first one is
wrong and the secondary is correct the algorithm never
looked up the secondary one. Please review my diff and feel free to
submit the patch onto your CVS.

If reading manual page resolv.conf(5) under nameserver and how the
algorithm should work the previous dns_lookup did not fully followed
that.

Regards

Imre Sunyi

20 years agoJoakim Tjernlund writes:
Eric Andersen [Sat, 24 Jan 2004 07:59:30 +0000 (07:59 -0000)]
Joakim Tjernlund writes:

Hi again

I have done some serious clenup of powerpc/elfinterp.c

- Make it look more like glibc.
- Performance modifictions.
- Fixed a few bugs in _dl_do_reloc(). These seem not to have affected
  anything, but this how glibc do it.

      Jocke

20 years agoSome minor debian junk I had sitting in my tree
Eric Andersen [Thu, 22 Jan 2004 01:44:25 +0000 (01:44 -0000)]
Some minor debian junk I had sitting in my tree

20 years agoSort all the files
Eric Andersen [Wed, 21 Jan 2004 23:45:14 +0000 (23:45 -0000)]
Sort all the files

20 years agoignore host binaries
Eric Andersen [Wed, 21 Jan 2004 23:30:52 +0000 (23:30 -0000)]
ignore host binaries

20 years agoSplit up syscalls.c, since it had grown to be quite large and ugly.
Eric Andersen [Wed, 21 Jan 2004 23:27:48 +0000 (23:27 -0000)]
Split up syscalls.c, since it had grown to be quite large and ugly.
 -Erik

20 years agoAdd poll
Eric Andersen [Wed, 21 Jan 2004 21:06:38 +0000 (21:06 -0000)]
Add poll

20 years agoMove poll to its own file, rather than being only half in its own file
Eric Andersen [Wed, 21 Jan 2004 20:27:25 +0000 (20:27 -0000)]
Move poll to its own file, rather than being only half in its own file

20 years agoMassive formatting cleanup, making it easier to parse
Eric Andersen [Wed, 21 Jan 2004 20:14:41 +0000 (20:14 -0000)]
Massive formatting cleanup, making it easier to parse

20 years agoErik is an idiot
Eric Andersen [Tue, 20 Jan 2004 21:36:23 +0000 (21:36 -0000)]
Erik is an idiot

20 years agosupport building the uClibc utils for the host system...
Eric Andersen [Tue, 20 Jan 2004 21:28:42 +0000 (21:28 -0000)]
support building the uClibc utils for the host system...

20 years agoAs noted by Doru Petrescu, there was a spurious line that was using
Eric Andersen [Tue, 20 Jan 2004 18:00:16 +0000 (18:00 -0000)]
As noted by Doru Petrescu, there was a spurious line that was using
_dl_debug_addr prior to checking if it was NULL.

20 years agoPaul Mundt writes:
Eric Andersen [Tue, 20 Jan 2004 17:44:17 +0000 (17:44 -0000)]
Paul Mundt writes:

The current behavior of the powerpc boot1_arch.h seems somewhat broken.
Currently room is made on the stack pointer for the link register, but the link
register is never actually pushed onto it. glibc bears the following comments:

    /* Call _dl_start with one parameter pointing at argc */
            mr      r3,r1
    /* (we have to frob the stack pointer a bit to allow room for
       _dl_start to save the link register).  */

followed by the -16 add to r1. Despite the fact that r1 is modified, the link
register is never actually pushed onto r1, thus the adjustment is completely
superfluous.

There's two possible fixes for this, either saving the link register in the way
that glibc does, or getting rid of the r1 adjustment. As I'm not sure if saving
the link register will actually break the _dl_boot2 callin, both options will
probably want to be played with.

The following bit of inline assembly in the attached patch builds cleanly for
me with gcc 3.3 on darwin, but I'm not able to test it any further beyond that.

20 years agoJoakim Tjernlund writes:
Eric Andersen [Tue, 20 Jan 2004 17:38:25 +0000 (17:38 -0000)]
Joakim Tjernlund writes:

> Is there anything I can do/provide that would assist in the solving of
> this problem on PowerPC? I'm still concerned about my lack of flash space...

Try this path. It fixes the recent ldso problems for me on my mpc860 board.
You should be able to use ldso without the dcbx patch to the kernel now.

Please report back.

 Jocke

20 years agoAdd __data_start needed by boehm garbage collector, in turn needed by libgcj.
Manuel Novoa III [Sat, 17 Jan 2004 14:45:55 +0000 (14:45 -0000)]
Add __data_start needed by boehm garbage collector, in turn needed by libgcj.

20 years agoPrevent runaway asserts, which can happen if there's a problem with
Manuel Novoa III [Sat, 17 Jan 2004 14:39:40 +0000 (14:39 -0000)]
Prevent runaway asserts, which can happen if there's a problem with
stdio or locales.

20 years agoWe need to link with libgcc.a when creating shared libs, in order to
Manuel Novoa III [Fri, 16 Jan 2004 18:52:54 +0000 (18:52 -0000)]
We need to link with libgcc.a when creating shared libs, in order to
avoid problems 'hidden symbol' problems.   Also handle -lfloat for the
soft-float arm case.

20 years agos/UCLIBC_HAS_MMU/ARCH_HAS_MMU/g
Eric Andersen [Fri, 16 Jan 2004 13:43:30 +0000 (13:43 -0000)]
s/UCLIBC_HAS_MMU/ARCH_HAS_MMU/g

20 years agoFixup selection of endianness, since select does not work on choice
Eric Andersen [Fri, 16 Jan 2004 13:38:10 +0000 (13:38 -0000)]
Fixup selection of endianness, since select does not work on choice
items, we have to declare what endianness cpus are capable of supporting
and work using dependancies.

20 years agoMinor updates from linux 2.6.1
Eric Andersen [Fri, 16 Jan 2004 12:49:06 +0000 (12:49 -0000)]
Minor updates from linux 2.6.1

20 years agoLink with libfloat if necessary.
Manuel Novoa III [Fri, 16 Jan 2004 00:46:39 +0000 (00:46 -0000)]
Link with libfloat if necessary.

20 years agoMake sure we remove -fomit-frame-pointer from the ldso build. It's
David McCullough [Fri, 16 Jan 2004 00:44:34 +0000 (00:44 -0000)]
Make sure we remove -fomit-frame-pointer from the ldso build. It's
inclusion will cause the loader to crash when jumping to the application.

The reason is due to the START macro having a "leave" instruction included
to fixup the stack before starting the app.

20 years agoInclude missing errno.h header (as noted by Alan Hourihane).
Eric Andersen [Wed, 14 Jan 2004 22:37:55 +0000 (22:37 -0000)]
Include missing errno.h header (as noted by Alan Hourihane).
Fix uninitialized pthread mutex used to lock the list of aligned
memory blocks.

20 years agoUse the gcc 3.4 -funit-at-a-time optimization when available
Eric Andersen [Wed, 14 Jan 2004 10:31:58 +0000 (10:31 -0000)]
Use the gcc 3.4 -funit-at-a-time optimization when available

20 years agoOk.. we need to link with libgcc.a for some archs. I haven't looked at
Manuel Novoa III [Wed, 14 Jan 2004 09:14:47 +0000 (09:14 -0000)]
Ok.. we need to link with libgcc.a for some archs.  I haven't looked at
all cases, but the archs I've looked at do build the needed routines as
PIC so we should be ok.

20 years agoAdd a dummy (always fails) wcsftime. This is needed for the libstdc++ locale
Manuel Novoa III [Wed, 14 Jan 2004 06:30:59 +0000 (06:30 -0000)]
Add a dummy (always fails) wcsftime.  This is needed for the libstdc++ locale
implementation to build.  Later this month I'll add a functional wcsftime.

20 years agoAdd another devfs subdir to the list of checked subdirs.
Manuel Novoa III [Tue, 13 Jan 2004 10:26:56 +0000 (10:26 -0000)]
Add another devfs subdir to the list of checked subdirs.

20 years agoFixup dated comments
Eric Andersen [Tue, 13 Jan 2004 05:58:00 +0000 (05:58 -0000)]
Fixup dated comments

20 years agoOops... Seems I forgot an else...
Manuel Novoa III [Mon, 12 Jan 2004 11:07:02 +0000 (11:07 -0000)]
Oops... Seems I forgot an else...

20 years agoAs mjn3 noticed, 64 bit arches should select UCLIBC_HAS_LFS
Eric Andersen [Fri, 9 Jan 2004 09:13:37 +0000 (09:13 -0000)]
As mjn3 noticed, 64 bit arches should select UCLIBC_HAS_LFS

20 years agoSome badly behaved arches fail to provide a PAGE_SIZE define
Eric Andersen [Tue, 6 Jan 2004 01:05:12 +0000 (01:05 -0000)]
Some badly behaved arches fail to provide a PAGE_SIZE define

20 years agoImplement res_search and res_querydomain based on netbsd code,
Eric Andersen [Sun, 4 Jan 2004 09:35:20 +0000 (09:35 -0000)]
Implement res_search and res_querydomain based on netbsd code,
but remove support for host aliases (the HOSTALIASES env variable)
which looks like a very bad idea.
 -Erik

20 years agoBump .deb version
Eric Andersen [Sun, 4 Jan 2004 09:32:14 +0000 (09:32 -0000)]
Bump .deb version

20 years agoFix paragraph seperation
Eric Andersen [Sun, 4 Jan 2004 05:49:55 +0000 (05:49 -0000)]
Fix paragraph seperation

20 years agospell 'pthread' properly
Eric Andersen [Sun, 4 Jan 2004 05:46:15 +0000 (05:46 -0000)]
spell 'pthread' properly

20 years agoFix grammar
Eric Andersen [Sun, 4 Jan 2004 05:34:35 +0000 (05:34 -0000)]
Fix grammar

20 years agoUpdate for release
Eric Andersen [Sun, 4 Jan 2004 05:32:21 +0000 (05:32 -0000)]
Update for release

20 years agoMy mistake. We do need some weak pthread_* funcs. But we were previously
Manuel Novoa III [Sun, 4 Jan 2004 01:29:01 +0000 (01:29 -0000)]
My mistake.  We do need some weak pthread_* funcs.  But we were previously
missing one in the glibc abi, and including a number of extras (which I
left in commented).
Unfortunately, this brings back the perl pthread bug.  Fixing the incorrect
weaks in libpthread wasn't sufficient to escape the shared loader bug.  :-(

20 years agoFix up the news a bit
Eric Andersen [Sat, 3 Jan 2004 11:17:11 +0000 (11:17 -0000)]
Fix up the news a bit

20 years agoUpdate for release
Eric Andersen [Sat, 3 Jan 2004 11:06:48 +0000 (11:06 -0000)]
Update for release

20 years agoBe sure to fully clean 2.6 kernel headers on 'make clean'
Eric Andersen [Sat, 3 Jan 2004 10:18:08 +0000 (10:18 -0000)]
Be sure to fully clean 2.6 kernel headers on 'make clean'

20 years agoAdd ldflags where needed
Eric Andersen [Sat, 3 Jan 2004 10:10:37 +0000 (10:10 -0000)]
Add ldflags where needed

20 years agoAnd just like that, clone is now fixed.... Previously the
Eric Andersen [Sat, 3 Jan 2004 08:49:04 +0000 (08:49 -0000)]
And just like that, clone is now fixed....  Previously the
error handling code was mostly broken.
 -Erik

20 years agooops
Eric Andersen [Sat, 3 Jan 2004 05:54:49 +0000 (05:54 -0000)]
oops

20 years agoMinor changes
Eric Andersen [Sat, 3 Jan 2004 05:52:34 +0000 (05:52 -0000)]
Minor changes

20 years agoA useful/useless little malloc time test
Eric Andersen [Fri, 2 Jan 2004 23:35:10 +0000 (23:35 -0000)]
A useful/useless little malloc time test

20 years agoPeter S. Mazinger writes:
Eric Andersen [Fri, 2 Jan 2004 23:34:13 +0000 (23:34 -0000)]
Peter S. Mazinger writes:

Hello Erik!

I have made some cosmetical changes to the files, removed the added
SCRT=-fPIC option from building the crt0.S file (but it is a requirement
to build them with -fPIC), and changed some comments. I have left the
ldso.c patch with PIE_SUPPORT ifdefs, but consider applying it w/o them
(see some earlier comment from PaX Team on this issue, as it is considered
a bug). To have it work correctly, you'll also need removing
COMPLETELY_PIC.
One thing is missing: PIE_SUPPORT should be usable only for i386 (for
now).

Also added the support for propolice protection (that works for me and
catches memcpy/strcpy attacks (but needs a special gcc version).

Thanks, Peter

20 years agoMinor update of the debian packaging stuff
Eric Andersen [Fri, 2 Jan 2004 23:24:17 +0000 (23:24 -0000)]
Minor update of the debian packaging stuff

20 years agoRemove remnants of the gcc wrapper
Eric Andersen [Fri, 2 Jan 2004 23:10:37 +0000 (23:10 -0000)]
Remove remnants of the gcc wrapper

20 years agoDo not make tests depend on CC, since we no longer build a wrapper.
Eric Andersen [Fri, 2 Jan 2004 23:07:57 +0000 (23:07 -0000)]
Do not make tests depend on CC, since we no longer build a wrapper.

20 years agoBump version, cleanup whitespace
Eric Andersen [Fri, 2 Jan 2004 12:02:24 +0000 (12:02 -0000)]
Bump version, cleanup whitespace

20 years agoCope with 2.6.x headers
Eric Andersen [Fri, 2 Jan 2004 11:59:28 +0000 (11:59 -0000)]
Cope with 2.6.x headers

20 years agorohde at soap dot adsl dot dk writes:
Eric Andersen [Fri, 2 Jan 2004 11:12:16 +0000 (11:12 -0000)]
rohde at soap dot adsl dot dk writes:

The macro to do some floating point checks in libc/sysdeps/linux/powerpc/setjmp.S is incorrect.
The following should fix it.

Same applies to uClibc/libc/sysdeps/linux/powerpc/__longjmp.S
Hope there aren't other files I've missed :)

20 years agorohde at soap dot adsl dot dk writes:
Eric Andersen [Fri, 2 Jan 2004 10:52:38 +0000 (10:52 -0000)]
rohde at soap dot adsl dot dk writes:

The macro to do some floating point checks in libc/sysdeps/linux/powerpc/setjmp.S is incorrect.
The following should fix it.

20 years agoFix __freadable and __fwritable... were using '~' instead of '!'. (ugh)
Manuel Novoa III [Fri, 2 Jan 2004 09:21:33 +0000 (09:21 -0000)]
Fix __freadable and __fwritable... were using '~' instead of '!'. (ugh)
Fix (hopefully) a potential problem with failed freopen() calls.  The
  fix isn't tested since I've been working on the replacement stdio
  core code which will go in after the next release.

20 years agoOops... temporarily support the old stdio code.
Manuel Novoa III [Fri, 2 Jan 2004 09:20:33 +0000 (09:20 -0000)]
Oops... temporarily support the old stdio code.

20 years agoThe checking Erik added seems to be unnecessary. Without it, I am not
Manuel Novoa III [Fri, 2 Jan 2004 08:47:22 +0000 (08:47 -0000)]
The checking Erik added seems to be unnecessary.  Without it, I am not
seeing any LTP failures.

20 years agoRedo the exec functions to comply with SUSv3.
Manuel Novoa III [Fri, 2 Jan 2004 08:44:58 +0000 (08:44 -0000)]
Redo the exec functions to comply with SUSv3.

20 years agoRewrite popen for SUSv3 compliance.
Manuel Novoa III [Fri, 2 Jan 2004 07:11:35 +0000 (07:11 -0000)]
Rewrite popen for SUSv3 compliance.
  Added a list of popen()'d to store pids and use waitpid() in pclose().
  Loop on waitpid() failure due to EINTR as required.
  Close parent's popen()'d FILEs in the {v}fork()'d child.
  Fix failure exit code for failed execve().

20 years agoMinor fix from upstream glibc.
Manuel Novoa III [Fri, 2 Jan 2004 07:01:13 +0000 (07:01 -0000)]
Minor fix from upstream glibc.

20 years agoKept the same approach, but rewrote the code for the most part.
Manuel Novoa III [Fri, 2 Jan 2004 07:00:29 +0000 (07:00 -0000)]
Kept the same approach, but rewrote the code for the most part.
Fixed some minor issues plus (as I recall) one SUSv3 errno case.

20 years agoFir errno return when the file/path is NULL or points to an empty string.
Eric Andersen [Wed, 31 Dec 2003 12:13:27 +0000 (12:13 -0000)]
Fir errno return when the file/path is NULL or points to an empty string.
 -Erik

20 years agoAdd extra checks for sigprocmask and rt_sigprocmask syscalls.
Eric Andersen [Wed, 31 Dec 2003 11:50:08 +0000 (11:50 -0000)]
Add extra checks for sigprocmask and rt_sigprocmask syscalls.
The rt_sigprocmask syscall has broken error handling in 2.4.x
kernels, while the sigprocmask syscall appears to get things
right.  Regardless we should be extra careful, and add these
checks.

20 years agoFix the manditory typo in my last minute untested fix.
Eric Andersen [Tue, 30 Dec 2003 10:57:16 +0000 (10:57 -0000)]
Fix the manditory typo in my last minute untested fix.
 -Erik

20 years agoMake sleep behave itself properly inthe presence of SIGCHLD
Eric Andersen [Tue, 30 Dec 2003 10:45:45 +0000 (10:45 -0000)]
Make sleep behave itself properly inthe presence of SIGCHLD

20 years agoRework malloc. The new default implementation is based on dlmalloc from Doug
Eric Andersen [Tue, 30 Dec 2003 10:40:49 +0000 (10:40 -0000)]
Rework malloc.  The new default implementation is based on dlmalloc from Doug
Lea.  It is about 2x faster than the old malloc-930716, and behave itself much
better -- it will properly release memory back to the system, and it uses a
combination of brk() for small allocations and mmap() for larger allocations.
 -Erik

20 years agoUpdate and restore malloc-simple. Slow as molasses, but trivially
Eric Andersen [Tue, 30 Dec 2003 01:41:14 +0000 (01:41 -0000)]
Update and restore malloc-simple.  Slow as molasses, but trivially
simple and releases memory immediately when asked to do so.
 -Erik

20 years agoFix a really dumb bug introduced in version 1.4 of this file (a patch for ipv6
Eric Andersen [Sun, 28 Dec 2003 08:25:16 +0000 (08:25 -0000)]
Fix a really dumb bug introduced in version 1.4 of this file (a patch for ipv6
support) which could cause things like EOF and read errors while reading
/etc/services to always return a TRY_AGAIN.  The perl test suite would alloc a
larger buffer and try again until all memory was exhausted.  When we get a read
error, or EOF, it means we didn't get what we wanted, and so we should return
an error.  Doing so fixes the failing perl 5.8.2 test.
 -Erik

20 years agoFix whitespace
Eric Andersen [Sun, 28 Dec 2003 07:41:24 +0000 (07:41 -0000)]
Fix whitespace

20 years agoremove unused file
Eric Andersen [Sun, 28 Dec 2003 00:25:31 +0000 (00:25 -0000)]
remove unused file

20 years agoHandle the app_fini stuff in exit without requiring atexit().
Manuel Novoa III [Sat, 27 Dec 2003 23:36:23 +0000 (23:36 -0000)]
Handle the app_fini stuff in exit without requiring atexit().
This avoids pulling in all the malloc/free code for a simple true/false app.

20 years agoFix a long-standing bug with pthreads. A couple of linuxthreads files
Manuel Novoa III [Sat, 27 Dec 2003 23:30:50 +0000 (23:30 -0000)]
Fix a long-standing bug with pthreads.  A couple of linuxthreads files
were including libc-lock.h which had a bunch of weak pragmas.  Also,
uClibc supplied a number of no-op weak thread functions even though
many weren't needed.  This combined result was that sometimes the
functional versions of thread functions in pthread would not override
the weaks in libc.

While fixing this, I also prepended double-underscore to all necessary
weak thread funcs in uClibc, and removed all unused weaks.

I did a test build, but haven't tested this since these changes are
a backport from my working tree.  I did test the changes there and
no longer need to explicitly add -lpthread in the perl build for
perl to pass its thread self tests.

20 years agooops
Eric Andersen [Mon, 22 Dec 2003 11:22:59 +0000 (11:22 -0000)]
oops

20 years agoimplement the worthless cuserid() function we claim to support.
Eric Andersen [Mon, 22 Dec 2003 11:19:49 +0000 (11:19 -0000)]
implement the worthless cuserid() function we claim to support.
This isn't in SuSv3, but is expected by at least some apps such
as emacs...

20 years agoIn unistd.h, getpgrp() is redirected to __getpgid(), but
Eric Andersen [Mon, 22 Dec 2003 10:47:04 +0000 (10:47 -0000)]
In unistd.h, getpgrp() is redirected to __getpgid(), but
we did not have a __getpgid().   Fix that.

20 years agoThat was awfully stupid.
Eric Andersen [Mon, 22 Dec 2003 09:17:19 +0000 (09:17 -0000)]
That was awfully stupid.

20 years agoUse 'sed -i -e' rather than 'sed -ie' since we don't want
Eric Andersen [Mon, 22 Dec 2003 08:56:37 +0000 (08:56 -0000)]
Use 'sed -i -e' rather than 'sed -ie' since we don't want
to leave backup files with "e" appended scattered all over
the place.