[arch-general] [arch-dev-public] makepkg.conf settings - Was:[signoff] pacman 3.3.0
Gerardo Exequiel Pozzi
vmlinuz386 at yahoo.com.ar
Sun Aug 2 23:59:22 EDT 2009
Allan McRae wrote:
> Dan McGee wrote:
>> Allan (and others)- I added a carbon copy version of makepkg.conf to
>> SVN. Do we want to establish (1) Arch LDFLAGS or (2) Arch integrity
>> check policy?
>>
>
> Here go my recommendations:
>
> 1) Arch LDFLAGS: -Wl,--hash-style=gnu -Wl,--as-needed
> There are very few problems with --as-needed these days and several
> other distros are using it to. You can always do an somthing like
> export LDFLAGS="" if it fails (much like is done when our CFLAGS cause
> issues. The -Wl,--hash-style=gnu will cause us not to have sysv
> hashes in our packages (we currently patch gcc to have both so
> building without makepkg will be unaffected).
>
>
For LDFLAGS agree with you, but maybe is secure to also add
-Wl,--no-undefined , so I guess the linked can report undefined symbols
in regular object files. (I can't remember now, but without this flag
and with --as-needed some packages not build, but this was before the
recent as-needed patch to binutils)
I think that is sane to add to CFLAGS -Wall (and with more reason with
latest gcc, and only produce more/equal build msg output), why this?
Many apps already setup this flag in Makefiles, but others no. Do you
remember the issue with mysql corruption because strict-aliasing? This
can be detected at build time if -Wall was present. Of course at this
time many packages are rebuilded, but only god knows what os these have
an strict-aliasing issue, and is problematic.
My two cents,
--
Gerardo Exequiel Pozzi ( djgera )
http://www.djgera.com.ar
KeyID: 0x1B8C330D
Key fingerprint = 0CAA D5D4 CD85 4434 A219 76ED 39AB 221B 1B8C 330D
More information about the arch-general
mailing list