uclibc-ng

Fork

Joakim Tjernlund writes:

Hi it is me again.

This is the latest ldso patch. the NEW weak symbol handling works now
with a little special handling in _dl_find_hash(). You get to chose
if you want the new or old handling :)

There was 2 missing _dl_check_if_named_library_is_loaded() calls in _dlopen().

I then disabled the _dl_check_if_named_library_is_loaded() in dl-elf.c since
it is rendundant.

Question, why does some _dl_linux_resolver(), like i386, have 2 calls
to _dl_find_hash()? I think that is wrong, isn't it?

I really hope you can check this out soon ...

Yet more fixups..

Done free path if it equals not_found
Remember to flose an fopened file

Another little touch up to avoid problems...

A better fix...

Fix the annoying "not a dynamic executable" problem when running ldd
on libuClibc

Cleanup whitespace and formatting

Avoid potential double free when locate_library_file() is used
on fully resolved names

Oops.. The custom stream funcs should only be compiled if wanted.

Based on a problem report and patch from Florian Schirmer, fix a compile
problem with mips -- contrary to my expectation gcc on mips _always_ defines
__PIC__ which breaks the assumption inherent in my code. This should fix
things up.

More todo list updates

Paul Mundt lethal at linux-sh dot org writes:

We can't use trapa #0xff on sh2, the only permissible trapa ranges are
32 to 63. On SH-2, we use trapa32 to issue a break in the same way that
trapa #0xff is used on SH-3/4. This behavior is implementation specific,
but is what is used in sh-ipl+g, linux, eCos, uITRON, etc. so we follow
suit here.

Add missing mips specific include/sys/user.h

Add our own copies of the include/scsi header files, per what glibc
does, rather than depending on the kernel header files.
-Erik

New stdio core. Should be more maintainable. Fixes a couple of bugs.
Codepaths streamlined. Improved performance for nonthreaded apps
when linked with a thread-enabled libc.
Minor iconv bug and some locale/thread related startup issues fixed.
These showed up in getting a gcj-compiled java helloworld app running.
Removed some old extension functions... _stdio_fdout and _stdio_fsfopen.

Fix bug from the syscall reorganization, detected by python's test_poll.py.
The #ifdef __NR_poll test was failing because it was done before any includes.
Hence, the emulation was always being used.

NOTE: The emulation fails a couple of tests in test_poll.py!

yet another todo item

mention valgrind/atexit annoyance

Update TODO list with some items for 1.0

Do not include the create_module syscall if it is not present

Eliminate the PIE support option, and simply support that all the time

Fix function prototype to match the official ELF standard hash function

Fixup header file naming

clean trailing whitespace

Rework file naming, aiming for at least a vague level of consistancy

oops, manditory forgotten patch

Fixup the definition of _dl_open, and move some bits back where they were,
till I properly finish the next step in my evil plan.

Nicolas Pitre writes:

Hello Erik,

My patch adding ARM assembly soft-float routines to gcc also changes the
default FPA float word ordering for the saner VFP word ordering which is
also the order anyone would expect on a little endian machine.

Problem is that uClibc curently hardwire floats to big endian (FPA) ordering
in all cases.

Please consider the attached patch to fix this problem.


Nicolas

Move _dl_printf and _dl_malloc to ldso, which is a more sensible
location for this stuff

Cleanup whitespace and formatting