| Age | Commit message (Collapse) | Author |
|---|
|
Kill old crt0/__uClibc_main.
This breaks ABI. All apps and toolchain needs to be recompiled.
All archs except x86 are now broken and need to be fixed as x86.
PPC will be fixed shortly by me, the rest is left to the arch maintainers.
|
|
'include/atomic.h' to add in new atomic operations for use by NPTL. There are multiple files for PowerPC and Sparc for 'atomic.h'. I will let those architecture maintainers choose the correct file. The files come from glibc in 'sysdeps/ARCH/bits'.
|
|
To actually use it, one must define _DL_DO_FINI_IN_LIBC globally.
That will also break ABI for apps linked with previous version of crtX.S.
They must be relinked with this version to work again.
|
|
|
|
|
|
|
|
as the flags for all calls to 'as'
|
|
The attached patch generalizes the use of PIE (all archs are brought in
sync that use/mention it: x86/ppc/frv) and makes use of it building the
target utils.
Tested on x86, ppc should be tested, frv uses -fPIE at one location, but
at another place -fpie, I don't know which is correct (could be both) and
misses the target addition in Config.in.
The test for ppc (requires the earlier sent crt-correction patch to work
correctly):
enable UCLIBC_PIE_SUPPORT, build uClibc and utils, check:
file ./utils/ldd, it should show shared object (instead of executable)
|
|
size and performance penalty to profiling applications this way, as well as
Heisenberg effects, where the act of measuring changes what is measured.
There are better tools for doing profiling, such as OProfile, that do not
require gcc to instrument the application code.
-Erik
|
|
-Erik
|
|
|
|
Some utilities, such as valgrind, have a legitimate reason to know the address
of the current brk. Since we know such utils will peek under our skirt, we
might as well give them what they expect and not use a gratuitously different
symbol name.
-Erik
|
|
with s/i686.get_pc_thunk.bx/get_pc_thunk_bx/g to make gcc 3.4 happy.
|
|
have it. It is used by the boehm gc, amoung other things.
|
|
|
|
'__kernel_old_dev_t'. And of course there is no good way to know
which is in use except checking linux/version.h. Grumble.
This is rather lame, but for now, define __kernel_old_dev_t to be
the same as __kernel_dev_t. This will want to be revisited soon.
-Erik
|
|
|
|
error handling code was mostly broken.
-Erik
|
|
|
|
|
|
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
|
|
being cleaned up.
|
|
used to generate the crti.S and crtn.S files. Since we don't use that
anymore, keeping the workaround makes no sense.
Furthermore, in most cases, SAFECFLAGS was not picking up all the
needed flags, causing crti.o and crtn.o to not be built PIC.
Which is very bad. Removing SAFECFLAGS and using CFLAGS fixes
that as well.
|
|
broke a couple of days ago. :-(
|
|
|
|
|
|
ln.patch:
* Define $(LN) as ln in Rules.mak.
* Change all occurrences of ln into $(LN).
* Change all constructs like (cd path && ln -sf foo/file file)
into $(LN) -sf foo/file path/file. The latter construct is
already used in a number of places so it should not be
an additional compatibility problem.
|
|
rm.patch:
* Define $(RM) as rm -f in Rules.mak and test/Rules.mak
(this is the same definition as gmake uses by default).
* Change all occurrences of rm and rm -f into $(RM).
|
|
Current uClibc contains only one fpu_control.h and it is i386 version.
This is a patch to use platform specific fpu_control.h. All new files
come from glibc 2.3.2. This patch is against 0.9.21 but also can be
applied to CVS as is.
|
|
|
|
on i386, at least, so seems like a good thing.
|
|
in registers.
NOTE: i386/mcount.S really needs to be rewritten. It currently won't
work for non-PIC builds.
|
|
|
|
|
|
|
|
|
|
|
|
should instead have been testing for '#ifdef __PIC__'. This resulted in
NON-PIC code getting mixed into the shared library. Oops!!!
-Erik
|
|
|
|
|
|
|
|
|
|
type of 'struct stat' and 'struct stat64' so they use consistant types.
This change is the result of a bug I found while trying to use GNU tar. The
problem was caused by our using kernel types within struct stat and trying to
directly compare these values with standard types. Trying an 'if (a < b)' when
'a' is an 'unsigned long' and 'b' is an 'int' leads to very different results
then when comparing entities of the same type (i.e. time_t values)....
Grumble. Nasty stuff, but I'm glad I got this out of the way now.
As a result of this fix, uClibc 0.9.17 will not be binary compatible with
earlier releases. I have always warned people this can and will happen.
-Erik
|
|
generate a crt0 and crt1 file. Most arches still need
to be updated to call __uClibc_start_main() rather than
__uClibc_main().
|
|
|
|
be weak so people won't need to fix their compilers
|
|
implementation such that gdb can actually debug signal handlers. Gdb behaves
much better now, for example, on multi-threaded apps.
-Erik
|
|
build a crt0.o and a crt1.o. crt1.o will support ctors and dtors if
such support is enabled. One more gratuitous toolchain support issue
is thereby removed...
-Erik
|
|
|
|
configurable, so people who do not need or want ctor/dtor support
can disable it and make their binaries a little bit smaller.
-Erik
|
