gh-100257: summarize_stats: Link failure kinds to GitHub search #100258
Conversation
By adding links to each row in the failure kind table, we can find where in the code that failure kind has been collected by the specialization code.
✅ Deploy Preview for python-cpython-preview canceled.
|
|
It looks like this:
Sample search result: array int. |
|
That link returns a page with "We couldn’t find any code matching 'repo:python/cpython path:Python/specialize.c SPEC_FAIL_SUBSCR_ARRAY_INT' ". Since this can produce misleading search results for branches other than main, it is likely to be counter productive for experimental stats and changes to the stat names. |
Hmm, it works for me.
That could be resolved since we know the hash of the Python that produced it, we could include that in the search URL, if GitHub search supports such a thing. That would make them "permanent" in a sense. IMHO, this is better than nothing despite that issue -- it is hard to know what these snippets of text refer to. Perhaps better (but obviously more work) would be to document them. |
This link? |
|
I'm not sure linking to the source for the macro is going to help clarify stuff. |
|
The link now leads me (iff I'm signed in to Github) to the line The PR has significant conflicts now. Should it be closed, or improved? |
|
I think we should give this sufficient time for the OP to comment before closing just in case (👋 @lpereira! Hope things are well with you!), but... a lot has changed since this was posted, including a major refactor of summarize_stats.py (so this would be a full rewrite, probably), and full help messages have been added to these tables. It's not as sophisticated as this, but it gets us closer to answering "what does this mean?" for the user of this output. |
|
Hi! Thanks for the ping, Mike! (Things are good here, BTW.)
I don't think I'm going to update this PR, and it seems people didn't find it that useful, so I think it's OK to close it.
…On Wed, Apr 10, 2024, at 8:16 AM, Michael Droettboom wrote:
I think we should give this sufficient time for the OP to comment before closing just in case (👋 @lpereira <https://github.com/lpereira>! Hope things are well with you!), but... a lot has changed since this was posted, including a major refactor of summarize_stats.py (so this would be a full rewrite, probably), and full help messages have been added to these tables. It's not as sophisticated as this, but it gets us closer to answering "what does this mean?" for the user of this output.
—
Reply to this email directly, view it on GitHub <#100258 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAADVGJSGQWHHNU2FSWTLRDY4UUSPAVCNFSM6AAAAAAS7BQUDGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANBXGM4TAOJTGM>.
You are receiving this because you were mentioned.Message ID: ***@***.***>
|
|
Thanks, @lpereira. Going to close this, then. |
By adding links to each row in the failure kind table, we can find where in the code that failure kind has been collected by the specialization code. Search is performed only in
specialize.cin thepython/cpythonrepo (the script is unaware of forks at the moment).