2012-05-09 Andreas Jaeger <aj@suse.de>
* nscd/nscd.c (run_modes): Make named enum, reorder so that
default is first entry.
(run_mode): Set type.
(main): Remove informal message about syslog.
(options): Fix typo.
[BZ #14053]
GCC 4.7 might remove consecutive calls to e.g. lrintf since
the assembler instructions are the same and GCC does not know
that the result is different depending on the rounding mode. For
SSE instructions, the control register is not available so there
is no way to inform GCC about this. Therefore the asms are marked
as volatile.
For a better integration of nscd with systemd, we should run nscd not as
daemon but in the foreground. A new --foreground option should be added.
2012-05-09 Alexandre Oliva <aoliva@redhat.com>
Andreas Jaeger <aj@suse.de>
* nscd/nscd.c (go_background): Replaced with...
(run_mode): ... this.
(RUN_FOREGROUND, RUN_DAEMONIZE, RUN_DEBUG): Add.
(options): Add -F --foreground.
(main): Implement it.
(parse_opt): Parse it.
[BZ #14083]
Fix warning when using strspn with -Wconversion:
$ gcc -Wconversion -O t.c
t.c: In function ‘main’:
t.c:8:7: warning: conversion to ‘long unsigned int’ from ‘int’ may change the sign of the result [-Wsign-conversion]
If the compiler and flags would select the hard-float ABI
then the ARM configure fragment will set HAVE_ARM_PCS_VFP.
This is later used by the ARM shlib-versions to select
the appropriately named dynamic linker.
The current tftp structure does not work when fortification is enabled.
Starting with gcc-4.5, more size checking was added to trigger these.
Older versions just didn't have enough information, so they returned -1
as the sizes.
First, the tu_stuff field is declared as 1 byte (when it's really an
arbitrary length C string), so attempting to strcpy() with it results
in crashes. This fails with _FORTIFY_SOURCE=1.
Second, even if we change that to [0] (since gcc does not allow flexible
array members in an union), gcc is not smart enough to see that they are
two overlapping flexible arrays (tu_stuff and tu_data), so it will still
trigger an abort with _FORTIFY_SOURCE=2. This is because it thinks that
tu_stuff is 0 bytes and tu_data comes after it.
Talking to upstream gcc, they don't seem terribly inclined to fix the
2nd issue, but even if they did, we still have plenty of 4.5 and 4.6
installs that would hit problems.
So, let's re-order with a few more anonymous structs & unions so that
the fields are laid out with a zero-length array always as the last
field. This seems to fix things with gcc-4.6, and the tftp-hpa pkg
continues to build & work.
URL: https://bugs.launchpad.net/ubuntu/+source/tftp-hpa/+bug/691345
URL: https://bugs.archlinux.org/task/28103
URL: https://bugs.gentoo.org/357083
URL: http://gcc.gnu.org/PR52944
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
tst-obprintf failed with GCC 4.7.
It turned out that this is the fault of GCC optimizing away the
following from malloc/mcheck.c:
/* We call malloc() once here to ensure it is initialized. */
void *p = malloc (0);
free (p);
gcc sees the malloc(0);free pair and removes it completely.
And now malloc is not properly initialized and we screw up if both
mcheck is used (via tst-obprintf) and MALLOC_CHECK_ is set (as it is in my
environment).
