[issue29137] Fix fpectl-induced ABI breakage

STINNER Victor added the comment: I created bpo-43250: "[C API] Depreate or remove PyFPE_START_PROTECT() and PyFPE_END_PROTECT()". By the way, I close this issue: there is no activity since 2019. -- status: open -> closed ___ Python tracker

[issue29137] Fix fpectl-induced ABI breakage

STINNER Victor added the comment: I propose PR 17231: "PyFPE_START_PROTECT() and PyFPE_END_PROTECT() macros are empty: they do nothing for one year (since commit 735ae8d), stop using them." -- nosy: +vstinner ___ Python tracker

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: Yes, they're intentionally retained as no-ops, so that it remains possible to load old extensions that were compiled against an fpe build and refer to those symbols. Is there a problem? -- ___

[issue29137] Fix fpectl-induced ABI breakage

Matthias Klose added the comment: reopening. this patch introduces two new symbols unconditionally, which were not defined for non-pyfpe builds before (PyFPE_counter and PyFPE_jbuf). -- nosy: +doko status: closed -> open ___ Python

[issue29137] Fix fpectl-induced ABI breakage

Change by Benjamin Peterson : -- resolution: -> fixed stage: patch review -> resolved status: open -> closed ___ Python tracker

[issue29137] Fix fpectl-induced ABI breakage

Benjamin Peterson added the comment: New changeset 735ae8d139a673b30b321dc10acfd3d14f0d633b by Benjamin Peterson (Nathaniel J. Smith) in branch 'master': bpo-29137: Remove fpectl module (#4789)

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: Ping -- anyone up for reviewing PR 4789? https://github.com/python/cpython/pull/4789 It's pretty straightforward, and I figure better to ping now and beat the end-of-month rush :-). Also, it has an autoconf refresh in it, so it's likely to

[issue29137] Fix fpectl-induced ABI breakage

Change by Serhiy Storchaka : -- nosy: +rhettinger, stutzbach, twouters ___ Python tracker ___

[issue29137] Fix fpectl-induced ABI breakage

Change by Nathaniel Smith : -- keywords: +patch pull_requests: +4689 stage: -> patch review ___ Python tracker ___

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: @Dima: > @njs: to point out that usefulness of this module is not just wishful > thinking. I just used it to locate, up to the line in a Python extension > module written in C, a bug in Sagemath (that has perhaps 20 FPU-using > extensions, some of them as

[issue29137] Fix fpectl-induced ABI breakage

Dima Pasechnik added the comment: PS. I would volunteer to fix it and maintain it, assuming I have some modest funding to support such an activity. (What precisely "it" should be, is another question). -- ___ Python tracker

[issue29137] Fix fpectl-induced ABI breakage

Dima Pasechnik added the comment: @njs: to point out that usefulness of this module is not just wishful thinking. I just used it to locate, up to the line in a Python extension module written in C, a bug in Sagemath (that has perhaps 20 FPU-using extensions, some of them as large as numpy).

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: Another option you might want to consider is proposing to add a proper fpu control flag setting/checking API to the math module. -- ___ Python tracker

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: Also fixing the abi issues that started this, and probably making an argument for why it makes sense for all of cpython's built-in float operations to check the fpu flags, and to do so using a weird longjmp-based mechanism that only some platforms support.

[issue29137] Fix fpectl-induced ABI breakage

Dima Pasechnik added the comment: @njs: Fixing this entails switching over from ieeefp to fenv, right? Looks doable, although not trivial. It might potentially be useful for various numerics, IMHO. -- ___ Python tracker

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: @Dima: are you volunteering to fix and maintain it? I can see why it's useful to have some way to get at the fpu flags, but I don't see how fpectl specifically helps with that issue, and fpectl has always been broken on x86-64. --

[issue29137] Fix fpectl-induced ABI breakage

Dima Pasechnik added the comment: While fpectl might be a bit rusty, this is a priceless tool in debugging situations, where one needs to identify components that do something wrong to FPU, see e.g. https://github.com/numpy/numpy/issues/9007 and the related

[issue29137] Fix fpectl-induced ABI breakage

Changes by Mark Dickinson : -- nosy: +mark.dickinson ___ Python tracker ___ ___

[issue29137] Fix fpectl-induced ABI breakage

Nick Coghlan added the comment: Ah, cool - I didn't know that Debian built with fpectl enabled by default. In that case, +1 for: - low maintenance ABI and API compatibility shims that are kept around indefinitely (since they leaked into the limited API/stable ABI) but don't actually do

[issue29137] Fix fpectl-induced ABI breakage

Nathaniel Smith added the comment: > At that point, does it actually make sense to provide the shims? Or should we > instead just add the deprecation warnings and say "Don't use the > --with-fpectl option, as it doesn't work properly, and breaks ABI > compatibility for extension modules built

[issue29137] Fix fpectl-induced ABI breakage

Nick Coghlan added the comment: - 3.4 is already in security-fix only mode so we can safely ignore it for this purpose - 3.5.3 is likely to be the last general bugfix release of 3.5, so we can probably skip that as well - that would mean the ABI compability shims would only go in 3.6.1 and

[issue29137] Fix fpectl-induced ABI breakage

Changes by Nick Coghlan : -- nosy: +ncoghlan ___ Python tracker ___ ___ Python-bugs-list

[issue29137] Fix fpectl-induced ABI breakage

New submission from Nathaniel Smith: It turns out that CPython built with --with-fpectl has a different ABI than CPython built without --with-fpectl (which is the default). Specifically, if you have an extension module built against a --with-fpectl CPython, and it uses the