| Age | Commit message (Collapse) | Author |
|---|
|
changes:
* Do not force -msmall-data if -fpic and -mlarge-data if -fPIC.
Instead, have both -fpic and -fPIC only select pic code generation,
and use whatever memory mode is specified (with -mlarge-data being
the implicit default).
* -Wstack-larger-than support.
* Disable stack frame checking in the prologue under OpenBSD, unless
you compile with -fstack-check, for consistency with other platforms.
* -mno-ieee option to disable IEEE mode.
ok kettenis@
|
|
|
|
actual cross support.
|
|
|
|
|
|
|
|
ok robert@
|
|
suggestion kettenis ok drahn
|
|
'looks reasonable' kettenis@
|
|
almost certainly breaks non-powerpc targets.
|
|
|
|
progress.
most of the smarts from drahn, ok drahn
|
|
to miod@ "has a history of being required to build security concerned
software".
ok guenther@, jsg@, robert@
|
|
|
|
|
|
|
|
removed that are now unnecessary.
|
|
okay millert@
|
|
|
|
prev_nonnote_insn() in push_frame_in_operand; ok miod@
|
|
some programs working correctly in tmux and screen.
From the readline changelog:
w. Fixed a bug in the callback read-char interface to make it work when a
readline function pushes some input onto the input stream with
rl_execute_next (like the incremental search functions).
ok deraadt
|
|
2004-08-10 Mark Mitchell <mark@codesourcery.com>
* expr.c (operand): Handle the "~", "-", and "!" operators applied
to bignums.
http://sourceware.org/cgi-bin/cvsweb.cgi/src/gas/expr.c.diff?cvsroot=src&only\
_with_tag=MAIN&r2=1.52&r1=1.51&f=u
Fixes the problem mentioned here:
http://llvm.org/bugs/show_bug.cgi?id=6528
ok jsg@, miod@
|
|
ok miod@
|
|
- disable unsafe builtins
- create a stub for -Wbounded
- move strict aliasing to -O3
|
|
simplify_shift_const(), and only do this on mips64 platforms, because we are
too late in the release cycle.
This change has the nice side effect of preventing a wrong overoptimistic
optimization on mips64, when masking bit 31 of a long; which in turns causes
python to misbehave.
Help and ok otto@
|
|
is intended to be used when compiling kernel code which will run on a
Loongson 2E or 2F processor, and inserts an explicit BTB clear operation
before every jump through a register (jr or jalr instructions), unless
that register is k0 or k1, or we are in .set noat.
This is a reliable, although aggressive, workaround for the misbehaviour
of the branch prediction engine of many Loongson 2F processors with regard
to the Branch Translation Buffer, for which an official errata has yet
to be published, and for which the Loongson suggested workaround apparently
only works due to pipeline side effects, but requires all the executable
kernel code to be located in CKSEG0/CKSEG1.
These changes are inspired by a backport of the Loongson binutils 2.19 patch
to binutils 2.18; the idea of doing an aggressive BTB clear is mine.
This causes no functional change to code compiled without the
-mfix-loongson2f-btb option.
``looks reasonable'' kettenis@
|
|
files or directories when applicable.
The inspiration and name of MACHINE_CPU come from NetBSD, although the way to
provide it to Makefiles is completely different.
ok kettenis@
|
|
Actually we should probably look further than just the immediate
predecessor here. ok miod@
|
|
when compiling boost 1.41.
From http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17327 via
Markus Hennecke <markus-hennecke at markus-hennecke dot de>.
miod@ tested and ok
|
|
|
|
are checked in.
|
|
|
|
|
|
in non-obj dir.
|
|
syntax errors found by mandoc(1), also required to fix the mandoc build;
ok jmc@
|
|
when checking out xenocara from a server running OpenBSD/amd64.
While processing RCS deltas, don't allocate twice as much memory as
needed when copying a line vector to a vector which has less lines.
Also, when switching back from a branch to trunk while searching an
RCS file for a revision, free the trunklines vector immediately after
lines saved in it have been copied back into the currentlines vector.
Somehow, these two changes together make the problem go away.
ok tobias@, "this has been a serious annoyance" sthen@, "sure" deraadt@
|
|
ok deraadt
|
|
2008-04-27 (2.8.7dev.9)
* pass a newline after the start of PRE-section in HTGopher.c to force
the first newline between records to be seen and cause the lines to
split (report by Ciprian Dorin Craciun) -TD
|
|
|
|
a backslash. Fix from Russ Allbery, the podlators maintainer.
|
|
up data after all allocations succeeded, free previously allocated
data if later allocation fails.
Prompted by parfait and based on a patch from jsg@; additional bits
from me and Claus Assmann of Sendmail. ok deraadt@
|
|
and hampered the inliner without anyone noticing.
fixes gcc bug #36367
reported by henning petersen
|
|
crash on certain invalid UTF-8 codes.
ok millert@ sthen@
|
|
information so this works on i386 as well.
ok robert@
|
|
|
|
implementation since random mmap breaks the default approach.
Note that gcc may not be compiled with PIE support or it will
break this implementation too.
|
|
value doesn't get truncated to int - on 64-bit architectures.
ok miod@, millert@, espie@
|
|
|
|
- disable -fident by default
- disable -fstrict-aliasing with -O2
- add support for -Wstack-larger-than-N (only for i386 and amd64 for now)
|
|
|
