3 failures in make check, cxx, gmp 5.1.1, Solaris 10

Marc Girod marc.girod at gmail.com
Mon Mar 18 10:14:00 CET 2013


On Mon, Mar 18, 2013 at 6:26 AM, Marc Glisse <marc.glisse at inria.fr> wrote:

> A grep in the sources does not show any explicit 'throw' in the library,
> only in the header, and those are the ones we are testing.

I was just summarizing comments read in various reports.
I also guess the throw may not be indirect, so that
grepping the sources is not a full proof?

> Well now it may print
> FAIL: t-do-exceptions-work-at-all-with-this-compiler
> which hopefully will be a good hint of what is going on.

Good. I believe it is worth getting this issue clarified, if as
it seems, it affects a large number of users across several
platforms and several versions of gcc.

I'll try to be back once I'll have build my compiler, eventually.
I'll build gdb as well. Any clue what drives to a truncated
core dump? I checked that I have: ulimit 'unlimited'.

BTW, in the build process, platform options set to CCFLAGS
are not pushed consistently to CXXFLAGS, which results
in link errors in the cxx tests (at the very end of a lengthy
process...) Shouln't they?

I had e.g. at one point in config.log (yes, I have tried various
combinations, with no much certainty):

CC=gcc -std=gnu99
CFLAGS=-O2 -pedantic -m64 -mptr64 -mcpu=v9
path= sparc64 generic

Well, that's a recipe for failure, and configure could detect it.


More information about the gmp-bugs mailing list