[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [bug-mes] tests and uldiv on mes wip-arm
From: |
Jan Nieuwenhuizen |
Subject: |
Re: [bug-mes] tests and uldiv on mes wip-arm |
Date: |
Thu, 16 May 2019 21:26:59 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/26.2 (gnu/linux) |
Danny Milosavljevic writes:
Hi,
> I'm in the process of fixing some of the remaining failing tests on ARM and
> found those:
>
> FAIL: lib/tests/scaffold/23-pointer-sub.c [uldiv]
> FAIL: lib/tests/scaffold/36-compare-arithmetic.c [uldiv] [legitimate]
>
> I have a few questions:
>
> * How should div.o have been included? It doesn't seem to be
> included, hence uldiv could not be resolved.
The tests (see build-aux/test-c.sh) 00-2x do not link with any library;
30-4x link with libc-mini.
Maybe lib/mes/div.c (in build-aux/configure-lib.sh) could/should be
moved from libmes to libc-mini...but that won't work with
23-pointer-sub.c.
What we could also do, is disable 23-pointer-sub.c and
36-compare-arithmetic.c for arm (in build-aux/check-mescc.sh), and
then re-add them as 50-pointer-sub.c and 36-compare-arithmetic.c so
that they come with libc. Lots of possibilities, now what's the
easiest/cleanest way...?
> * The first one is dividing by 1. Should we detect that in the
> compiler and optimize it away?
Hmm, if that gets us past this that may be a good hack. Or maybe we can
remove the statement in 23-pointer-sub.c that produces this div?
> Also, a LOT of the other failing tests segfault. It would be nice to
> have a program in mes that finds out where that is in the C source (if
> the debug info is good enough for that).
We should be able to compile with some debug info, try using the -g flag
on mescc. mescc-tools's blood-elf will enable gdb to show a stack
trace.
Greetings,
janneke
--
Jan Nieuwenhuizen <address@hidden> | GNU LilyPond http://lilypond.org
Freelance IT http://JoyofSource.com | AvatarĀ® http://AvatarAcademy.com