Menu
▾
▴
quantlib-users
|
From: Luigi B. <lui...@gm...> - 2026-04-02 10:29:43
|
Hi all, a release candidate for QuantLib 1.42 is available at < https://github.com/lballabio/QuantLib/releases/tag/v1.42-rc>. Python wheels are also available at < https://test.pypi.org/project/QuantLib/1.42rc0/>, and a C# NuGet package is at <https://int.nugettest.org/packages/QuantLib/1.42.0-rc>. If you have some time, please try it out and report any issues on GitHub (or here on the mailing list). Thanks! Luigi |
|
From: Dirk E. <ed...@de...> - 2026-04-10 13:16:29
|
On 2 April 2026 at 12:29, Luigi Ballabio wrote: | Hi all, | a release candidate for QuantLib 1.42 is available at < | https://github.com/lballabio/QuantLib/releases/tag/v1.42-rc>. Python | wheels are also available at < | https://test.pypi.org/project/QuantLib/1.42rc0/>, and a C# NuGet package is | at <https://int.nugettest.org/packages/QuantLib/1.42.0-rc>. If you have | some time, please try it out and report any issues on GitHub (or here on | the mailing list). Thanks! All good from here on the Debian side. Built (and uploaded) QuantLib yesterday, followed up with QuantLib-SWIG (for just Python) now. I did once again have to carry the 'make Python 3.10 the minimum' patch forward as I had for the last two (?) cycles. I presume you keep it 3.8 for a reason? If not a gentle nudge to adjust :) The patch basically is still the same as https://sources.debian.org/src/quantlib-swig/1.41-1/debian/patches/upgrade-minimum-python-version-3.10.patch but will get slightly changed line numbers when '1.41.99' (as we cannot append '-rc' in our version sorting) pops up in due course. Cheers, Dirk -- dirk.eddelbuettel.com | @eddelbuettel | ed...@de... Sponsor me for Tour de Shore 2026! Donate at https://www.pledgereg.com/536435 More about the ride at https://dirk.eddelbuettel.com/blog/2026/04/03#sponsor_tour_de_shore_2026 |
|
From: Luigi B. <lui...@gm...> - 2026-04-13 12:55:18
|
Hi Dirk, as usual thanks for the support. > I did once again have to carry the 'make Python 3.10 the minimum' patch forward as I had for the last two (?) cycles. Meaning that you tried not applying it and the build failed? I thought it was no longer necessary if Debian picked up SWIG 4.4.1 (the upgrade to 4.4.0 made it necessary, 4.4.1 should have fixed it). > I presume you keep it 3.8 for a reason? Well, since it doesn't require us to create extra wheels (the ones we create with the stable ABI work with all versions) it's supposed to be zero-cost and possibly make some people happier. But if it still requires you to patch the build, I'll probably drop it next time. Luigi On Fri, Apr 10, 2026 at 3:17 PM Dirk Eddelbuettel <ed...@de...> wrote: > > On 2 April 2026 at 12:29, Luigi Ballabio wrote: > | Hi all, > | a release candidate for QuantLib 1.42 is available at < > | https://github.com/lballabio/QuantLib/releases/tag/v1.42-rc>. Python > | wheels are also available at < > | https://test.pypi.org/project/QuantLib/1.42rc0/>, and a C# NuGet > package is > | at <https://int.nugettest.org/packages/QuantLib/1.42.0-rc>. If you have > | some time, please try it out and report any issues on GitHub (or here on > | the mailing list). Thanks! > > All good from here on the Debian side. Built (and uploaded) QuantLib > yesterday, followed up with QuantLib-SWIG (for just Python) now. I did once > again have to carry the 'make Python 3.10 the minimum' patch forward as I > had > for the last two (?) cycles. > > I presume you keep it 3.8 for a reason? If not a gentle nudge to adjust :) > The patch basically is still the same as > > > https://sources.debian.org/src/quantlib-swig/1.41-1/debian/patches/upgrade-minimum-python-version-3.10.patch > > but will get slightly changed line numbers when '1.41.99' (as we cannot > append '-rc' in our version sorting) pops up in due course. > > Cheers, Dirk > > -- > dirk.eddelbuettel.com | @eddelbuettel | ed...@de... > > Sponsor me for Tour de Shore 2026! Donate at > https://www.pledgereg.com/536435 > More about the ride at > > https://dirk.eddelbuettel.com/blog/2026/04/03#sponsor_tour_de_shore_2026 > > > _______________________________________________ > QuantLib-users mailing list > Qua...@li... > https://lists.sourceforge.net/lists/listinfo/quantlib-users > |
|
From: Dirk E. <ed...@de...> - 2026-04-13 13:27:50
|
Hi Luigi, On 13 April 2026 at 14:54, Luigi Ballabio wrote: | Hi Dirk, as usual thanks for the support. | | > I did once again have to carry the 'make Python 3.10 the minimum' patch | forward as I had for the last two (?) cycles. | | Meaning that you tried not applying it and the build failed? Yes. There were really two problems here. First, once a patch is placed into debian/patches/ in the sources and referenced in debian/patches/series, it is always applied (using a system called `quilt`, which AFAIK is a general tool from outside Debian). Usually that 'just works' as the sources do not change. Here it failed, I believe the last and this time. So with the patch NOT applied the build then failed. And am I guessing correctly that there even was/is the intent to automate this as there are Python/setup.py and Python/setup.py.in and I may well be doing something wrong if I have to patch both? | I thought it was no longer necessary if Debian picked up SWIG 4.4.1 (the upgrade to | 4.4.0 made it necessary, 4.4.1 should have fixed it). I would not know -- you are the swig expert here. | > I presume you keep it 3.8 for a reason? | | Well, since it doesn't require us to create extra wheels (the ones we | create with the stable ABI work with all versions) it's supposed to be | zero-cost and possibly make some people happier. But if it still requires | you to patch the build, I'll probably drop it next time. Can setup.py somehow pick up how/where it builds and could we condition on that? There must be a way to support both 'very conservative Python 3.8' for manywheels and all that as well as our (and presumably other people's) use on newer systems that do not go back as far? In my case 'manually' adjusting is not a big deal. If we could make the patch auto-apply it would be even easier. Could setup.py 'read a file' we patch (and that does not change) or 'listen to an env var' (that I could set) ? Dirk -- Dirk Eddelbuettel | ed...@de... | http://dirk.eddelbuettel.com |
|
From: Luigi B. <lui...@gm...> - 2026-04-13 14:15:04
|
Ok, I see the problem — apt is still giving you SWIG 4.4.0: see < https://buildd.debian.org/status/package.php?p=swig&suite=sid>. And quite simply, 4.4.0 had a bug. I would have thought sid had 4.4.1 at this point, given that it was released last December. Upgrading to that SWIG version would fix the build. If you can get hold of the maintainer, maybe you can nudge them? Luigi On Mon, Apr 13, 2026 at 3:27 PM Dirk Eddelbuettel <ed...@de...> wrote: > > Hi Luigi, > > On 13 April 2026 at 14:54, Luigi Ballabio wrote: > | Hi Dirk, as usual thanks for the support. > | > | > I did once again have to carry the 'make Python 3.10 the minimum' patch > | forward as I had for the last two (?) cycles. > | > | Meaning that you tried not applying it and the build failed? > > Yes. There were really two problems here. First, once a patch is placed > into > debian/patches/ in the sources and referenced in debian/patches/series, it > is > always applied (using a system called `quilt`, which AFAIK is a general > tool > from outside Debian). Usually that 'just works' as the sources do not > change. Here it failed, I believe the last and this time. So with the > patch > NOT applied the build then failed. > > And am I guessing correctly that there even was/is the intent to automate > this as there are Python/setup.py and Python/setup.py.in and I may well be > doing something wrong if I have to patch both? > > | I thought it was no longer necessary if Debian picked up SWIG 4.4.1 (the > upgrade to > | 4.4.0 made it necessary, 4.4.1 should have fixed it). > > I would not know -- you are the swig expert here. > > | > I presume you keep it 3.8 for a reason? > | > | Well, since it doesn't require us to create extra wheels (the ones we > | create with the stable ABI work with all versions) it's supposed to be > | zero-cost and possibly make some people happier. But if it still > requires > | you to patch the build, I'll probably drop it next time. > > Can setup.py somehow pick up how/where it builds and could we condition on > that? There must be a way to support both 'very conservative Python 3.8' > for > manywheels and all that as well as our (and presumably other people's) use > on > newer systems that do not go back as far? In my case 'manually' adjusting > is > not a big deal. If we could make the patch auto-apply it would be even > easier. Could setup.py 'read a file' we patch (and that does not change) > or > 'listen to an env var' (that I could set) ? > > Dirk > > -- > Dirk Eddelbuettel | ed...@de... | http://dirk.eddelbuettel.com > |
|
From: Dirk E. <ed...@de...> - 2026-04-13 17:13:44
|
On 13 April 2026 at 16:14, Luigi Ballabio wrote: | Ok, I see the problem — apt is still giving you SWIG 4.4.0: see < | https://buildd.debian.org/status/package.php?p=swig&suite=sid>. And quite | simply, 4.4.0 had a bug. Ahh. Easy enough. That explains it. | I would have thought sid had 4.4.1 at this point, given that it was | released last December. Upgrading to that SWIG version would fix the | build. If you can get hold of the maintainer, maybe you can nudge them? I will file a courtesy bug report gentlu nudging the maintainer(s). [ Some time passes. ] Done, see #1133688. Dirk -- Dirk Eddelbuettel | ed...@de... | http://dirk.eddelbuettel.com |
