Re: Help with neovim segfault on mips64el (Loongson specific?)

James Cowgill <[email protected]> Fri, 2 Feb 2018 14:32:12 +0000

Newsgroups	gmane.linux.debian.ports.mips
Message-ID	<[email protected]>

Hi,

On 28/01/18 19:52, James McCoy wrote:
> Hi all,
> 
> Recently neovim has been segfaulting during tests in both the neovim[0] and
> neovim-qt[1][2] builds.  This appears to be specific to Loongson
> hardware, as the build[3] on mipsel-sil-01.d.o (Octeon) worked.  I'm
> also unable to reproduce the problems on eller (another Octeon).
> 
> [0]: https://buildd.debian.org/status/fetch.php?pkg=neovim&arch=mips64el&ver=0.2.2-2&stamp=1516772635&raw=0
> [1]: https://buildd.debian.org/status/fetch.php?pkg=neovim-qt&arch=mips64el&ver=0.2.8-2&stamp=1516592952&raw=0
> [2]: https://buildd.debian.org/status/fetch.php?pkg=neovim-qt&arch=mips64el&ver=0.2.8-3&stamp=1517149409&raw=0
> [3]: https://buildd.debian.org/status/fetch.php?pkg=neovim&arch=mips64el&ver=0.2.2-2&stamp=1516818666&raw=0
> 
> Does anyone have some insight into possible issues or would be willing
> to look into the problem?

The segfault happens inside libjemalloc which means this is very likely
caused by #843926 ("jemalloc uses a hard coded page size detected during
build"). It only fails on Loongsons because they have 16K pages whereas
the Octeons have 4K pages.

James

signature.asc (application/pgp-signature, 833 B)