Prepare for upcoming *.txt -> *.adoc bulk rename #1944
Description
There is a proposed patch series to rename the *.txt files in Documentation/ to *.adoc.
Where *.txt files are available, people frequently run various commands to parse and consume them, including tooling for static websites as this one. Which means that such a rename breaks that tooling.
I agree that this would have been a nice feature to add at the beginning of the development of the documentation, but I fear that it is too late to make an incompatible change now. However, as opposed to increasing Git's safety stance, the convenience of a few who want to have syntax highlighting in editors based on the file extension apparently is important enough to merit just such an incompatible change.
This means that script/update-docs.rb will need to be adapted once the tooling here breaks. Or before it. If someone volunteers.
Metadata
Metadata
Assignees
Labels
Type
Projects
Milestone
Relationships
Development
Participants
Activity
erikdendekker commentedon Feb 19, 2025
Just like to point out that the link to the latest release notes (Shown in monitor image on the front page and download page) is now broken because of this. It points to https://raw.githubusercontent.com/git/git/master/Documentation/RelNotes/2.48.1.txt while it should point to https://raw.githubusercontent.com/git/git/master/Documentation/RelNotes/2.48.1.adoc
dscho commentedon Feb 19, 2025
That's sad. Do you have time to work on this? I currently lack the time.
dscho commentedon Feb 24, 2025
@erikdendekker you can thank @aeiouaeiouaeiouaeiouaeiouaeiou over in #1954 for fixing this.
CodingGuidelines& friends todocs/#1965To1ne commentedon Mar 5, 2025
@dscho It seems the patch series landed. And 2.49 is around the corner. How hard is the script breaking now? Shall we pick this up?
dscho commentedon Mar 5, 2025
Yep.
Yep.
For comparison:
And for the record, these lines indicate a regression thanks to the Meson changes:
@To1ne If you have the time and the motivation; I do not.
To1ne commentedon Mar 11, 2025
I just noticed we have to make sure not to break permalinks like
https://git-scm.com/docs/git-clone#Documentation/git-clone.txt-code--filterltfilter-specgtcode. There'sgit-clone.txtin that anchor. 😱dscho commentedon Mar 11, 2025
Well spotted, I missed that!
This
.txt->.adocrename seems even less thought-through than I originally believed.To1ne commentedon Mar 12, 2025
Hah, there are more issues. Derp.
https://lore.kernel.org/git/CAEiLEbOZ7vGE6U69sf5nK+G86zaeAMRTrjaCr=rF2JU1H1p8ww@mail.gmail.com/
Hi Git Community,
The link to the release notes for v2.48.1 on the git-scm downloads
page doesn't seem to be working.
It links to: https://raw.githubusercontent.com/git/git/master/Documentation/RelNotes/2.48.1.txt
It looks like the master branch now uses '.adoc' extension since this
commit: git/git@1f010d6
Using either of these URLs loads the release notes correctly:
dscho commentedon Mar 12, 2025
@To1ne don't you love the impulsive response "This ain't our problem". As if the millions of links out there to Git's release notes that were in such a consistent format for over a decade so that people had come to rely on it, sometimes in commits (that are immutable) weren't Git's responsibility. I'm speechless.
dscho commentedon Mar 12, 2025
As for https://git-scm.com/downloads, this seems to have been adjusted in #1954, though.
jnavila commentedon Mar 12, 2025
Well, Hyrum's law doesn't really apply here, because they went directly to breaking the interface. As Junio himself says, the RelNotes were never meant to be asciidoc.
dscho commentedon Mar 12, 2025
Oh, but Hyrum's law totally applies here, because it's not up to maintainers to decide what users expect, and just wishing away users' expectations is a strategy that's always going to fail, every single time.
Junio can say that
RelNoteswere never meant to be AsciiDoc as long and as often as he wants, he can chant "Kumbaya" and do a backwards handspring while doing so if he so wishes, and it still would not change users' expectations one bit, not even in the slightest.The fact that they went directly to breaking the interface very much implies that they violated Hyrum's law, big time, whether willfully or obliviously is unclear. I wonder whether there would have been smarter decisions that could have been made.
jnavila commentedon Mar 12, 2025
Is this change compatible with the processing of translations where I would also rename files
*.txt→*.adoc?dscho commentedon Mar 12, 2025
@jnavila #1973 does nothing to the
_l10npart of the code, it only touchesexpand_contentandindex_doc, leavingexpand_l10nandindex_l10n_docunchanged. Which means that the same change is needed there, too, if you follow upstream's decision to rename those files and make it your own.dscho commentedon Mar 12, 2025
@jnavila I adapted the script in #1976 so that it will also handle
git-html-l10nwhen that contains.adocfiles instead of.txtones.What took me the longest time was to make the
bundle exec make allthing work, it really wanted to fail with "Could not find gem 'polib' in locally installed gems.". Eventually, after runningapt-getidentically to how it is run in therubyworkflow, it finally, finally worked (but every iteration of the step, it rebuilt thepo4a-stamprule, which took a looooong time). Even after that, I was quite puzzled how to update my localgit-html-l10nbranch. How do you do that? Do you always manually replace the files? It was quite cumbersome for me.To1ne commentedon Mar 13, 2025
@dscho Sorry for missing that. I somehow knew I should fix those too, but then I forgot. But to be honest, I don't fully understand why those codepaths are so different?
I seem to have been completely unaware of the workflow to follow to generate localized docs. I didn't realize how much more complex the workflow is to get localized docs. I hope we've covered all now for the upcoming 2.49 release.
dscho commentedon Mar 13, 2025
@To1ne to be fully transparent, I had looked at the repository name,
git-html-l10n, spotted the "html" and thought that the.adocfiles didn't play a role.Only when I looked more closely after @jnavila raised his concerns did I spot that AsciiDoctor is still used in that code path.
What took the longest time was to create the local commit in the
git-html-l10nclone that does have those.adocfiles.I hope that this is all that's needed in git-scm.com to react to that bulk rename. However, my intuition tells me that this kind of disruptive change is a gift that will keep on giving. We'll have to see. Based on what I observed, there is no help to be expected from those who caused the disruption.
To1ne commentedon Mar 13, 2025
@dscho The only help I'm expecting is them saying "hey X is broken" when they discover something isn't working. And if we make everything work flawlessly, we'll hear nothing.
dscho commentedon Mar 13, 2025
Well, we all say that X is broken. Twitter was much better, and certainly not seriously under-staffed.
Yep. Story of my life.
To1ne commentedon Mar 14, 2025
You're right: #1978