GMP-5.0.3, t-bdiv test fail

[Pierre]

In chapter 6(.14) I've got a fail on "t-bdiv" test (When executing 'make check)

To fix this issue I simply ignore the "fix minor error part": sed -i 's/np + dn, qn/& - dn/' mpn/generic/dcpi1_bdiv_q.c

I didn't found any reference on gmp site for version 5.0.3 on this specific fix (Maybe is it a "artifact" of older version of LFS)?

[Gilles Espinasse]

Did you have gcc-4.3.2 (not 4.3.1 or 4.3.3). This is a compiler version that have know issue with gmp-5 as stated in ​http://gmplib.org/#STATUS

gmp-5.0.3 has a fix for the issue announced in july with 5.0.2 but is broken in some changes as stated in ​http://gmplib.org/list-archives/gmp-announce/2012-January/000031.html

5.04 should happen soon and have in NEWS

A bug in the test suite causing buffer overruns during "make check"
sometimes leading to subsequent malloc crashes, has been fixed.

Changelog on source tree even say

2012-02-09 , 5.0.4 is released

but is not yet announced

[Pierre]

Yes, compiled with gcc 4.3.2. I've seen that there was issues with this version but as I said I didn't have to "fix" or "apply" any patch to get a "successful" build/test. I just wanted to underline that if I apply the fix provided in LFS 6.14 (7.1 SVN) it fail. So it is maybe a good idea to remove that fix from the book no?

[Matthew Burgess]

Replying to Zophar:

Yes, compiled with gcc 4.3.2. I've seen that there was issues with this version but as I said I didn't have to "fix" or "apply" any patch to get a "successful" build/test.

Well, if you're not following the book (the book uses GCC-4.6.2, not GCC-4.3.2) then there's not much we can do to help you really. Is there a particular reason you're using such an old version of GCC?

[Pierre]

Hum sorry I meant 4.6.2... I followed the book stricly...

[Gilles Espinasse]

You are right that the sed should have been removed with gmp-5.0.3 upgrade as that fix was commited in july in the tree ​http://gmplib.org:8000/gmp-5.0/rev/538dfce27f41

With sed against 5.0.3, another -dn is added to the line and that should be a wrong change.

[Matthew Burgess]

Thanks Gilles. I'll fix that up tonight and try to work out why jhalfs didn't bail out on me! Thanks for the report as well, Zophar.

[Matthew Burgess]

Fixed in r9738.