Bug 641166 – sys-libs/glibc-2.26-r3 [Rap - linux-standalone/amd64] experimental_malloc = 'yes'

• Product: Gentoo/Alt
• Status: RESOLVED NEEDINFO

[Description Steven Trogdon 2017-12-15 17:26:10 UTC]

On both gentoo and the subject rap after

ebuild =sys-libs/glibc-2.26-r3 configure

the relevant config.log contains

experimental_malloc='yes'

I don't know what affect this has on gentoo but on the subject rap there are situations where one gets

*** Error in `python2.7': free(): invalid next size (fast): 0x0000000005229c10 ***
and
*** Error in `python2.7': corrupted double-linked list: 0x000000000783f3d0 ***

and other errors which go away when glibc is built as

export EXTRA_ECONF="--disable-experimental-malloc" && emerge -1 glibc

The Errors are documented at the comments above https://github.com/cschwan/sage-on-gentoo/issues/494#issuecomment-351818507

I'm not sure exactly how to provide a use case that demonstrates the issue.

[Comment 1 Benda Xu 2017-12-19 07:54:57 UTC]

(In reply to Steven Trogdon from comment #0)
>
> I'm not sure exactly how to provide a use case that demonstrates the issue.

Thanks for the report, is it possible to isolate the related (test) code from sage?

[Comment 2 Benda Xu 2017-12-19 07:59:43 UTC]

Toolchain team,

From glibc-2.26, a new experimental_malloc has been introduced and enabled by configure by default.  It seems to break sage, python and ecl related.  Could you please comment on this bug to hint which should be fixed?

Benda

[Comment 3 Sergei Trofimovich (RETIRED) 2017-12-19 08:21:08 UTC]

I think we'll need more info or something simple to reproduce. Is there a gentoo package that exhibits failure?

[Comment 4 Steven Trogdon 2018-01-24 17:30:52 UTC]

FWIW I no longer see this issue with glibc-2.26-r5. I'm not sure if there was a change in glibc or something on my prefix or both.

[Comment 5 Benda Xu 2018-01-24 20:09:09 UTC]

Thanks for the update.  Feel free to reopen it if the issue rises again.