Dear mentors,

I am looking for a sponsor for speedcrunch 1.1.0-1.

I am the upstream maintainer. Following Bastian Germann's (bage) advice, this
is a QA upload with the Maintainer field left unchanged.

* Homepage: https://www.speedcrunch.org/
* License: GPL-2+, CC0-1.0, BSD-2-clause
* Section: math
* Packaging: https://salsa.debian.org/debian/speedcrunch

Package details and source:
https://mentors.debian.net/package/speedcrunch/

dget -x https://mentors.debian.net/debian/pool/main/s/speedcrunch/speedcrunch_1.1.0-1.dsc

This updates SpeedCrunch to 1.1 and simplifies the packaging to use
upstream CMake defaults and the bundled manual. It also compiles
translations, replaces the DocBook man page with roff, removes obsolete
patches and XPM handling, and fixes the AppStream release entry.

The package builds successfully with sbuild on unstable.
Source and binary Lintian checks report no messages. The installed
application and man page work.

This is my first Debian packaging contribution. Feedback is welcome.

Regards,
Helder Correia

Hi Helder,

I am sponoring this without the docs and tests change.
We are very interested in Debian to build from source and run as much of
upstream's tests as we can.

The tests run fine as-is, so it would be a package regression to disable
them.

Let's see whether this upload will bring us reproducibility.

Thanks,
Bastian

Hi Bastian,

Thank you for sponsoring the upload and reviewing the packaging.

I had already traced the remaining reproducibility issue to resource ordering in sphinxcontrib-qthelp. The builder doesn’t globally sort the resource list, so directory traversal order can change the generated .qhp file and then the Qt Help databases embedded in the executable. SOURCE_DATE_EPOCH addresses timestamps, but not this ordering.

That was also one reason I disabled manual generation. Upstream ships the static manual, which I regenerate just before every release, so using that copy avoids repeating the generation and introducing this variability.

I understand Debian’s preference for rebuilding from source. I’ve attached a patch to address the ordering issue while keeping manual generation enabled. The workaround produced identical Qt Help files across multiple controlled local runs with different traversal orders.

Could you please include it?

Thanks again,
Helder