Fix commit in gh-publish-only

[abachma2]

When trying to publish changes made in from the last few PRs merged, I got an error message that

git commit -m "Generated master for `git log source -1 --pretty=short --abbrev-commit`" && git push --force upstream master
fatal: ambiguous argument 'source': both revision and filename
Use '--' to separate paths from revisions, like this:
'git <command> [<revision>...] -- [<file>...]'

This change should fix this error, removing the ambiguity in the source branch and the source directory and allow us to publish the changes.

[gonuke]

source should not be redundant at this point, so we should look at this more carefully. By design (if memory serves correctly), this make target should:

• fetch the upstream
• checkout the master branch (which should have no source directory)
• reset the HEAD (which does something subtle but important - perhaps removing the source dir from the current working dir)
• copy all the content from the BUILDDIR
• git add all the copied content
• delete the BUILDDIR
• git commit

At the time of commit there should not be any source directory to cause ambiguity. If it is there, it can sometimes be an indication of earlier stages not functioning properly?

[abachma2]

The steps in the Makefile are:

make clean
make html
git fetch $(GH_UPSTREAM_REPO)
git checkout -B $(GH_PUBLISH_BRANCH) $(GH_UPSTREAM_REPO)/$(GH_PUBLISH_BRANCH)
git reset HEAD
rsync -a $(BUILDDIR)/* .
rsync -a $(BUILDDIR)/.* .
git add -f `(cd $(BUILDDIR); find . -type f; cd ..)`
rm -rf  $(BUILDDIR)
git commit -m "Generated $(GH_PUBLISH_BRANCH) for `git log $(GH_SOURCE_BRANCH) -1 --pretty=short --abbrev-commit`" && git push --force $(GH_UPSTREAM_REPO) $(GH_PUBLISH_BRANCH)
	git checkout $(GH_SOURCE_BRANCH)

$(BUILDDIR) is ./gh-build, ${GH_PUBLISH_BRANCH) is master, $(GH_SOURCE_BRANCH) is source, and $(GH_UPSTREAM_REPO) is upstream

So it does do all of the steps you described. The master branch does have a directory names source. I don't remember having issues with this when we last updated the website back in 2021, so maybe this is an update in git?

[gonuke]

I don't think it's related to changes in git. I recognize this failure from way back, and I'm not sure how/when the source directory ended up in our master branch. I don't think it's supposed to be there... I'm digging through old commits to try to understand (or figure out why I remember this incorrectly...)

[abachma2]

Have you had time to look into/think about this any?

[gonuke]

Have you had time to look into/think about this any?

I tried to figure out if we need/expect the source directory to be there in the final published version of the website. My recollection is that we should not need that directory and that it's not supposed to be there. However, when I look through old commits, that directory has been present for a very long time - I think.

I'm trying to figure out how to confirm this, however, because we force push to main when we publish, so if the source directory was added recently, it may appear to be long standing in the git logs.

I guess the real question is whether anything on the website links into that directory and whether it would break the website if that directory was gone?

[gonuke]

I searched a recent local build of the website and didn't find any links that referred to this source folder.

[abachma2]

I can always try to remove the source directory from the main branch, then try to build and publish to see if that fixes it. We shuold be able to undo any changes that does.

[gonuke]

I'm pretty sure that deleting source and force pushing the master branch should be safe and consistent with what I thought was the right structure anyway.

[gonuke]

P.S. We should rename master to main at some point

[munkm]

@abachma2 how about we open another PR deleting the source folder from the master branch?

[abachma2]

I just tried:

• checking out master branch
• removing source directory
• checking out source branch
• building via make gh-publish
  I still got the fatal error described above, and the source branch reappeared in the master branch.

[gonuke]

I just tried:

• checking out master branch
• removing source directory
• checking out source branch
• building via make gh-publish
  I still got the fatal error described above, and the source branch reappeared in the master branch.

It looks like this changed in #303 and maybe it was not quite correct...

[gonuke]

I may have approved it without understanding carefully enough what was going on

[gonuke]

See PRs #346 and #345 to repair this to its originally intended state.

[abachma2]

The changes in #345 and #346 fixed what this PR addressed, so this PR is being closed.