Re: Bug#914814: spades: FTBFS with jemalloc/experimental

On Sat, 5 Jan 2019 11:05:38 +0200 Faidon Liambotis <paravoid@debian.org> wrote:
> Therefore attached is a patch that addresses this using the je_rallocx
> API. It also deals with the deprecation of the stats.cactive statistic
> (as of jemalloc 4.0), using the stats.active instead.
>
> The patch should be 3.6.0-compatible as well and can go in anytime and
> ahead of a potential jemalloc transition. It is lightly tested (i.e.

> builds and runs the test suite).

Dear Faidon,

Thank you for your patch. I was unable to get the tests to complete locally. Via an upload to experimental I confirmed that they do hang on the build machines as well: https://buildd.debian.org/status/fetch.php?pkg=spades&arch=amd64&ver=3.13.0%2Bdfsg-2%7E0jemalloc5&stamp=1547148149&raw=0

I see a couple paths forward:

1. The patch is evolved so that the tests pass

2. jemalloc is rolled back from 5 to 3 as it breaks an established package and the archive is starting to freeze for Buster

3. Same as 2, but a jemalloc5 package is introduced for packages that are compatible.

My goal is that Spades is not dropped from Buster, but looking at the schedule (and the length of the NEW queue) I am afraid that this may happen.

Obviously method 1 would be ideal, if it can be accomplished in time.