Hi Adrian! Adrian Bunk <[email protected]> writes:
> Control: tags 1150278 + patch > Control: tags 1150278 + pending > > Dear maintainer, > > I've prepared an NMU for gudhi (versioned as 3.13.0+dfsg-1.1) and > uploaded it to DELAYED/2. Please feel free to tell me if I should > cancel it. On the contrary, I'm extremely grateful – see below. > * dh_python3 --no-ext-rename (Closes: #1150278) > > This fixes loading the Python module, and therefore also the > autopkgtest. Fantastic! This bug was an embarassment rooted in me accidentally testing from the build tree. I was sure I would have to patch upstream's build scripts to set rpaths to fix it, so I'm absolutely delighted that you found a much simpler fix for me. > * Enable parallel building again. > > Looking at build times when it was disabled, it tripled build times on > the typical quad-core buildds. The riscv64 buildds have 4 slow cores and > 16 GB RAM, I'd hope that this will reduce the build time there from 29 hours > to around 10 hours. Good call. If I recall correctly, the issue was some some buildd where parallel builds tried to use far more than 4 threads, with little memory available. If that's no longer the case, then that's great. > If Ubuntu or anyone else still has problems today due to underresourced > buildds, then limiting parallelism based on actually measured memory > usage would be an option. Agree. > * Architecture: any, and remove workarounds for 32-bit architectures > that are no longer supported by python3-pot > > Getting rid of manual architecture lists removes the need to maintain > them. Yep, that was next on my list after I made the change to POT. Thanks again so much – this really helped in a period of time where I'm finding it difficult to find big enough chunks for spare time to fix issues like #1150278 in a timely manner. I'll integrate your diff into the package's Salsa repository ASAP. Best, Gard
signature.asc
Description: PGP signature
