Argument Clinic: Touch source file if any output file changed

[larryhastings]

BPO	20396
Nosy	@ncoghlan, @larryhastings, @zware, @serhiy-storchaka

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/larryhastings'
closed_at = None
created_at = <Date 2014-01-26.14:02:21.527>
labels = ['type-feature', 'expert-argument-clinic']
title = 'Argument Clinic: Touch source file if any output file changed'
updated_at = <Date 2015-02-25.15:25:31.345>
user = 'https://github.com/larryhastings'

bugs.python.org fields:

activity = <Date 2015-02-25.15:25:31.345>
actor = 'serhiy.storchaka'
assignee = 'larry'
closed = False
closed_date = None
closer = None
components = ['Argument Clinic']
creation = <Date 2014-01-26.14:02:21.527>
creator = 'larry'
dependencies = []
files = []
hgrepos = []
issue_num = 20396
keywords = []
message_count = 3.0
messages = ['209315', '209318', '209323']
nosy_count = 4.0
nosy_names = ['ncoghlan', 'larry', 'zach.ware', 'serhiy.storchaka']
pr_nums = []
priority = 'low'
resolution = None
stage = 'needs patch'
status = 'open'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue20396'
versions = []

Linked PRs

• gh-64595: Argument Clinic: Touch source file if any output file changed #104152
• gh-64595: Fix write file logic in Argument Clinic #104507
• gh-64595: Fix regression in file write logic in Argument Clinic #106449
• [3.12] gh-64595: Fix regression in file write logic in Argument Clinic (#106449) #106452

[larryhastings]

A minor bug, but still one that could cause confusion. It's already bitten me once.

If I change the code generator in clinic.py, then --make everything, files using the "file" destination often won't change. All the changes to the generated code will go in the generated file. And since the UNIX build process ignores those generated files, it doesn't rebuild the source file. I'd have to use "--make -f".

This won't just happen to me though. If you update your trunk and run "make clinic" this could easily happen to you too.

I propose that Clinic handle this as follows: if the generated file changes, but the source file does not, it "touches" the source file. (Technically I'll probably just rewrite the source file, as I already have code that does that.)

[serhiy-storchaka]

It should first touch source file and then generated file. So that the date of generated file is always later the date of source file.

[ncoghlan]

With Serhiy's modification, sounds like a reasonable plan to me.

[erlend-aasland]

Resolved, taking Serhiy's and Nick's remarks into account, with:

• gh-64595: Argument Clinic: Touch source file if any output file changed #104152

[erlend-aasland]

My patch was faulty; make clinic now touch pretty much every file with clinic code. So, gh-104152 actually made stuff worse than the previous status quo.

[erlend-aasland]

Resolved with:

• gh-64595: Fix write file logic in Argument Clinic #104507

[erlend-aasland]

#104507 introduced a regression which in practice makes clinic unusable. Clearly, the current ClinicExternalTest test is not sufficient; we need to be able to catch errors as serious this in CI.

[hugovk]

Triage: the fix PR has been merged, can this be re-closed?

[erlend-aasland]

Triage: the fix PR has been merged, can this be re-closed?

No, the fix introduced a regression, so it had to be reverted :( I'll see if I can find time in the coming weeks to fix this.

[erlend-aasland]

I'm not sure this is worth fixing in Argument Clinic. It is a limitation with how we've set up the build system. A more modern build system would detect the #include dependencies automatically. We can chose to fix this in the Makefile (or just wait until a new build system is in place).