setting ac_exeext=$EXEEXT after L1340 in configure.ac
use another variable than EXEEXT; for example EXE_SUFFIX
other workarounds?
My gut feel regarding these is that I'd really not like to add more EXEEXT hacks, so I'd like to avoid 1). 2) should be ok, given that no-one else are depending on EXEEXT (cc. @hroncok).
I am not entirely sure what's my input here supposed to be.
I just suspected a third-party consumer of our build system would like to receive a heads-up when I proposed a rename of one of the configure/Make variables.
Anyway, as I feared, the renaming broke CI, so I'm reverting it. I suggest instead to amend Dan Amelang's patch to use ac_exeext iso. EXEEXT.
I amended Dan's patch to use ac_exeext in Codespaces, and can confirm that that approach works for WASI.
OTOH, most of the macros in autoconf-archive does actually use EXEEXT, and not ac_exeext, implying that if we should land python/cpython-devcontainers#30, we'd possibly run into a similar issue with another macro at a later point1. Perhaps we should just work around this in configure.ac by making sure ac_exeext is set to the same as EXEEXT.
Footnotes
yet another reason to switch to a more modern build system ↩
Activity
erlend-aasland commented on Oct 18, 2024
cc. @damelang, author of AX_C_FLOAT_WORDS_BIGENDIAN.
erlend-aasland commented on Oct 18, 2024
Another workaround, could be to move the float checks to before our EXEEXT hacks, but who knows what will break next time.
pythongh-125698: Replace EXEEXT with EXE_SUFFIX
hroncok commented on Oct 18, 2024
I am not entirely sure what's my input here supposed to be.
gh-125698: Replace EXEEXT with EXE_SUFFIX (#125699)
Revert "pythongh-125698: Replace EXEEXT with EXE_SUFFIX (python#125699)"
pythongh-125698: Restore EXEEXT in configure and Make
erlend-aasland commented on Oct 20, 2024
I just suspected a third-party consumer of our build system would like to receive a heads-up when I proposed a rename of one of the configure/Make variables.
Anyway, as I feared, the renaming broke CI, so I'm reverting it. I suggest instead to amend Dan Amelang's patch to use ac_exeext iso. EXEEXT.
gh-125698: Restore EXEEXT in configure and Make (#125758)
erlend-aasland commented on Oct 20, 2024
I amended Dan's patch to use ac_exeext in Codespaces, and can confirm that that approach works for WASI.
erlend-aasland commented on Oct 20, 2024
See python/cpython-devcontainers#30 for a proposed patch.
erlend-aasland commented on Oct 21, 2024
OTOH, most of the macros in autoconf-archive does actually use EXEEXT, and not ac_exeext, implying that if we should land python/cpython-devcontainers#30, we'd possibly run into a similar issue with another macro at a later point1. Perhaps we should just work around this in configure.ac by making sure ac_exeext is set to the same as EXEEXT.
Footnotes
yet another reason to switch to a more modern build system ↩
pythongh-125698: Autoconf: Sync EXEEXT and ac_exeext
pythongh-125698: Autoconf: Sync EXEEXT and ac_exeext
gh-125698: Autoconf: Sync EXEEXT and ac_exeext (#125995)
pythongh-125698: Autoconf: Sync EXEEXT and ac_exeext (pythonGH-125995)
pythongh-125698: Autoconf: Sync EXEEXT and ac_exeext (pythonGH-125995)
erlend-aasland commented on Oct 26, 2024
Landed on the following solution:
Automerge enabled for the backports. Closing this as completed.
[3.12] gh-125698: Autoconf: Sync EXEEXT and ac_exeext (GH-125995) (#1…
[3.13] gh-125698: Autoconf: Sync EXEEXT and ac_exeext (GH-125995) (#1…
pythongh-125698: Replace EXEEXT with EXE_SUFFIX (python#125699)
pythongh-125698: Restore EXEEXT in configure and Make (python#125758)
pythongh-125698: Autoconf: Sync EXEEXT and ac_exeext (python#125995)