[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#408850: mksh build problems

unblock 421518 by 408850
reassign 408850 gcc
retitle 408850 using flags "-fwhole-program --combine" broken

Hi all,

thanks to Steve Langasek I now know that the “conflicting prototypes” issues
for the mksh package appear due to use of the gcc flags “-fwhole-program
--combine” which, supposedly, have gcc treat it as one single translation
unit and optimise better. I suspect that this doesn't handle redeclaration
well; it seems to be a gcc bug.

Interestingly enough, the very same flags cause problems when using the
FORTIFY_SOURCE definitions on SuSE (and I think Fedora) – it fails on a
line saying something like “if (read(fd, buf, size) < 0)” with a bunch
of weird error messages; Pascal Bleser, whom I've added to the Cc: of
this eMail can surely provide them to you if that's desired. (Thanks to
Marcus “darix” Rückert for helping to track this down as well.)

I have taken these facts as arguments against using the flags at all and
changed upstream mksh to not test for and use them at all any more. A new
mksh minor release (R29d) was put out; I'll have my sponsor upload the
upgraded debian package shortly.

This means: unblock 421518 by 408850, since it no longer depends on the
fix (so that 421518 can be closed by an upgraded package), change 408850
to be a gcc bug now that we, thanks to vorlon, know the case.

I believe no one can invent an algorithm. One just happens to hit upon it
when God enlightens him. Or only God invents algorithms, we merely copy them.
If you don't believe in God, just consider God as Nature if you won't deny
existence.		-- Coywolf Qi Hunt

Reply to: