On 20 June 2024 at 12:16, Bo YU wrote: | hi, | | On Mon, Jun 17, 2024 at 7:16 AM Bo YU <tsu.y...@gmail.com> wrote: | > | > Hi, | > | > On Mon, Jun 17, 2024 at 6:41 AM Dirk Eddelbuettel <e...@debian.org> wrote: | > > | > > | > > Hi Paul, | > > | > > Thanks for the prompt and detailed reply. | > > | > > On 16 June 2024 at 16:13, Paul Gevers wrote: | > > | Hi Dirk, | > > | | > > | On 16-06-2024 2:42 p.m., Dirk Eddelbuettel wrote: | > > | > I may need a hgand with riscv64. | > > | | > > | That's normally a question to the porters, in CC now, so they can have a | > > | look. | > > | | > > | > The 1.34-1 revision needed some build | > > | > changes I had done poorly in such a way that the -O0 no longer applied to | > > | > some arches, this has been fixed in 1.34-2 so armel, armhf, i386 are good. | > > | > But riskv64 still times out. | > > | | > > | Ack. | > > | | > > | > Can we expand the build-time window from the | > > | > (arguably already large) value? | > > | | > > | Not that I know of. | > > | | > > | > Or can we (worst case) turn riskv64 builds | > > | > off? | > > | | > > | That's up to you as a maintainer, but this should be last resort [1]. | > > | Don't forget to request for removal of the existing riscv64 binaries if | > > | you go this route. Please be aware of [2] if you aren't already. | > > | > > True true, and I think I had to pull this 'safety value' once or twice before | > > with challenging / large package. I will re-read [1] and [2] and ponder. | > > | > > riscv64 porters: I would of course also love to hear if you can offer any | > > advice. The package is a tricky one as it contains (a lot of) heavily | > > templated C++ code that is autogenerated via Swig for these Python | > > bindings. The compilation of that one file is tricky. | > > | | I am always trying to build it on my local Unmatched boards with many | attempts. But unfortunately, they all failed so far and each building | will last > 1d.:( | | But this does not mean it does not work on riscv. On sg2042, it can be built: | ``` | Build Architecture: riscv64 | Build Type: binary | Build-Space: 1463468 | Build-Time: 21213 | Distribution: unstable | Host Architecture: riscv64 | Install-Time: 88 | Job: /home/vimer/ftbfs/quantlib-swig/quantlib-swig_1.34-2.dsc | Lintian: warn | Machine Architecture: riscv64 | Package: quantlib-swig | Package-Time: 21368 | Source-Version: 1.34-2 | Space: 1463468 | Status: successful | Version: 1.34-2 | -------------------------------------------------------------------------------- | Finished at 2024-06-19T21:08:10Z | Build needed 05:56:08, 1463468k disk space | ``` | I am looking at other optimization methods, but this will take more time again.
Thank you! It is a (very) big package, but it has not grown much lately. Not being able to build may lead to auto-removal which is bad, excluding an architecture is also not good. Not clear what the least bad move is here... Dirk | BR, | Bo | | | | > Okay, I will have a look at this package. | > | > After a quick look, there are a lot of architecture-related build | > flags here, so I might start there first. | > | > BR, | > Bo | > > Best, Dirk | > > | > > | Paul | > > | | > > | [1] https://release.debian.org/testing/rc_policy.txt : Packages must be | > > | supported on as many architectures as is *reasonably* possible. | > > | (Emphasis mine). | > > | [2] https://lists.debian.org/debian-devel/2022/09/msg00105.html | > > | [DELETED ATTACHMENT OpenPGP_signature.asc, application/pgp-signature] | > > | > > -- | > > dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org | > > -- dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
