Gmp 6.1 in core-updates

Gmp 6.1 in core-updates

[Andreas Enge]

Hello,

I just had a quick look at gmp-6.1.0, to which I updated in core-updates,
and which causes failures on arm.

When trying to build from hand with "make check -j 4", I got the following:
make[4]: *** No rule to make target '../libgmp.la', needed by 't-bswap'.  Stop.
make[4]: *** Waiting for unfinished jobs....
Apparently, the checks were started before the library was compiled.

Well, that is not what we do. But it raises suspicions whether a parallel
build works.

Then I tried "make -j 4", which completed, followed by "make -j 4 check".
One of the tests fails:
FAIL: reuse
See tests/mpz/test-suite.log

The test passes using "make check". So I suggest to disable parallel checks
for gmp and will (try to) commit a corresponding patch.

Andreas

Re: Gmp 6.1 in core-updates

[Ludovic Courtès]

Andreas Enge <andreas@enge.fr> skribis:

> I just had a quick look at gmp-6.1.0, to which I updated in core-updates,
> and which causes failures on arm.
>
> When trying to build from hand with "make check -j 4", I got the following:
> make[4]: *** No rule to make target '../libgmp.la', needed by 't-bswap'.  Stop.
> make[4]: *** Waiting for unfinished jobs....
> Apparently, the checks were started before the library was compiled.
>
> Well, that is not what we do. But it raises suspicions whether a parallel
> build works.

This is weird, but indeed, not what gnu-build-system does.

> Then I tried "make -j 4", which completed, followed by "make -j 4 check".
> One of the tests fails:
> FAIL: reuse
> See tests/mpz/test-suite.log

Isn’t it again an instance of presumed overheating, as discussed in
<http://bugs.gnu.org/21620>?

These tests are purely computational AFAIK, so I don’t see how running
tests in parallel could possibly lead to failures.

WDYT?

Thanks,
Ludo’.

PS: Forget my other message on this subject, which I sent before reading
    yours.