This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] sim: make sure to include strsignal prototype
On Sat, 24 Mar 2012, Mike Frysinger wrote:
> > Did you "make check-sim RUNTESTFLAGS=--target_board=cris-sim"?
> > The warning message indicates the absence of that option or the
> > wrong baseboard.
>
> using those flags makes most run. the linker segfaults when working on
> helloaout.c, so that can't be good:
>
> $ cris-elf-gcc ../../sim/testsuite/sim/cris/c/helloaout.c \
> -O2 -L/usr/local/src/gnu/sourceware/build/build-cris/ld -sim3 \
> -Wl,-mcrisaout -sim -Ttext=0 -lm -o helloaout.c.x -Bld
> collect2: ld terminated with signal 11 [Segmentation fault], core dumped
Ouch. I'd be happy to process a ld PR for this, or I'll get to
it eventually. Not that supporting a.out is of any particular
interest, but it shouldn't pollute the test-results.
> > (Arguably the *sim* test-suite should know set that option by
> > itself, but at least it's consistent with other runnable
> > dejagnu-based test-suites.)
>
> no other target needs this.
You mean "when you run the test-suite with a target compiler,
the only other sim target with C test-cases doesn't need this". :)
I see bfin had a few C test-cases added the last year, nice.
> i did some work about two years ago specifically
> to avoid having to mess with dejagnu just to run `make check-sim`.
I guess you missed fixing CRIS then. :)
I'm not overexcited about the crusade making "sim" a special
case, when it works using the same dejagnu spell as anything
else. So _no_ adding compiler flags to the individual
test-cases for the globally needed options.
> looks like the difference are just compiler flags ?
Linker flags, pointing to the right library, stuff like that...
See the baseboard file. Different options for *-linux-gnu than
for *-elf though. (Yes, the sim can run trivial cris*-linux-gnu
-static programs.)
brgds, H-P