gitmeta.tex not found

[mcdittmar]

I was just working to install the workflows into the HighEnergyObsCoreExt repository and encountered a problem with this file.

First, the repo is looking at the most recent ivoatex (commit d772edd), pretty sure anyway.
The build was failing at \include gitmeta.tex because it couldn't find it.
It does generate fine when building on the command line.

It looks like the file is supposed to be auto-generated during the build (rather than included in the repository).

I've added a \IfFileExists clause around the input line, so it doesn't cause a problem for the build, but I'd prefer it works.

Let me know what simple detail I've missed!
Mark

[msdemlei]

On Thu, Mar 13, 2025 at 02:34:48PM -0700, Mark Cresitello-Dittmar wrote: Let me know what simple detail I've missed!

gitmeta.tex is generated by the Makefile *if* you tell it to (and you should do that if you input gitmeta.tex). The background is explained in ivoatexDoc sect 3.8 <https://ivoa.net/documents/Notes/IVOATexDoc/20230627/NOTE-ivoatexDoc-1.4-20230627.html#tth_sEc3.8>. Very shortly, the fix is to have SOURCES = $(DOCNAME).tex gitmeta.tex (and whatever else) in the Makefile.

[mcdittmar]

Very shortly, the fix is to have

SOURCES = $(DOCNAME).tex gitmeta.tex

That's why this is baffling me, our Makefile has that (line 22).
The gitmeta.tex file is generated and the info is in the PDF when run on the command line, but not in the Git workflow.

[msdemlei]

On Fri, Mar 14, 2025 at 05:52:11AM -0700, Mark Cresitello-Dittmar wrote: That's why this is baffling me, our Makefile has that (line 22). The gitmeta.tex file is generated and the info is in the PDF when run on the command line, but not in the Git workflow.

Ah! Yeah, we're catching this now, but now well enough. Anyway, I'm pretty sure this is because git is missing in the container. Could you see if make update-ci git push fixes the problem? I'd sure hope it does :-)

[mcdittmar]

container. Could you see if

make update-ci
git push

fixes the problem? I'd sure hope it does :-)

I executed this, but have not pushed it.
The command replaces the .github/workflow/[build|preview].yml files.
The only differences are:

• doc_name: field is blank
• on:push:branches: = master rather than main. This repository uses 'main'

Both of these are necessary changes to make before committing.
BTW: the instruction block which comes up from the command says nothing about updating the doc_name field before doing the commit & push.. it probably should.. right? The branch name depends on the repository.

[msdemlei]

On Fri, Mar 14, 2025 at 08:11:19AM -0700, Mark Cresitello-Dittmar wrote: I executed this, but have not pushed it. The command replaces the .github/workflow/[build|preview].yml files. The only differences are: * doc_name: field is blank * on:push:branches: = master rather than main. This repository uses 'main'

That will certainly not fix it. Can I call in @gmantelet? He knows *a lot* more about the CI integration than I do.

[mcdittmar]

That will certainly not fix it. Can I call in @gmantelet? He knows
a lot more about the CI integration than I do.

I wonder if this might be involved:
The workflow runs 'make biblio ${{ env.doc_name }}-draft.pdf'

• the biblio thread does not include gitmeta.tex in the dependency path
  • biblio: ${DOCNAME}.bbl
  • ${DOCNAME}.bbl: ${DOCNAME}.tex ivoatex/ivoabib.bib ivoatexmeta.tex

So that thread would do the build without gitmeta.tex getting made.
The biblio step creates the ${DOCNAME}.pdf so maybe that isn't getting regenerated during the ${{ env.doc_name}}-draft.pdf phase.

That might be fixed if the .bbl step used ${SOURCES} ?
I'm not to sure what the biblio thread does exactly, but I'd guess we'd want it to scan all the source files (though probably not the gitmeta.tex since it will definitely not have any citations

[mcdittmar]

So that thread would do the build without gitmeta.tex getting made. The biblio step creates the ${DOCNAME}.pdf so maybe that isn't getting regenerated during the ${{ env.doc_name}}-draft.pdf phase.

That might be fixed if the .bbl step used ${SOURCES} ? I'm not to sure what the biblio thread does exactly, but I'd guess we'd want it to scan all the source files (though probably not the gitmeta.tex since it will definitely not have any citations

On my command line, it does rebuild it, but the workflow log has:

latexmk -pdf HighEnergyObsCoreExt
Rc files read:
  /etc/LatexMk
Latexmk: This is Latexmk, John Collins, 31 Jan. 2024. Version 4.83.
Latexmk: Nothing to do for 'HighEnergyObsCoreExt.tex'.
Latexmk: All targets (HighEnergyObsCoreExt.pdf) are up-to-date

pdflatex ivoatex/draft-background.tex

[gmantele]

The problem seems to come from the fact that the rule gitmeta.tex is never called either in ivoatex/Makefile or in the CI actions.
I tried to add an explicit call to this rule in the preview action (in preview.yml) in my fork:

run: make gitmeta.tex biblio ${{ env.doc_name }}-draft.pdf

It seems to work.

However, @msdemlei may have a cleaner solution to solve this. It is strange that this rule is never called anywhere in the ivoatex/Makefile, except when it is added inside SOURCES in the document's Makefile. Or should it be called only from the CI?

[gmantele]

The difference between the build CI and the preview CI is that the first one runs a simple make while the second runs make biblio ${{ env.doc_name }}-draft.pdf. make will run everything inside SOURCES which is not the case of the preview CI. Maybe we should add SOURCES among the tasks to run in the preview CI. What do you think @msdemlei ?

[msdemlei]

On Thu, Mar 20, 2025 at 03:49:12AM -0700, Grégory Mantelet wrote: gmantele left a comment (ivoa-std/ivoatex#153) inside `SOURCES` which is not the case of the `preview` CI. Maybe we should add `SOURCES` among the tasks to run in the `preview` CI. What do you think @msdemlei ?

Hm, I see. The actual root problem is that $(DOCNAME).bbl does not depend on $(SOURCES), but it clearly should. So, we ought to change $(DOCNAME).bbl: $(DOCNAME).tex ivoatex/ivoabib.bib ivoatexmeta.tex into $(DOCNAME).bbl: $(SOURCES) ivoatex/ivoabib.bib ivoatexmeta.tex I think I didn't do that because I once had a plan to trick $(DOCNAME).bbl into SOURCES and have automatic bibliography building even without latexmk. But the way things turned out, I'd hope that should just work. Having said that, let's also drop biblio from make biblio ${{ env.doc_name }}-draft.pdf Since we know install latexmk into the preview container, we don't need explicit bibliography management any more. since $(DOCNAME)-draft.pdf depends on $(DOCNAME).pdf, all should just run fine (actually, even without the first fix, but the first fix is valuable nevertheless). Would you do the change? Or should I?

[gmantele]

You seem to have a perfectly clear idea on how to do that. So, if ok with you, I let you do it.

[msdemlei]

On Thu, Mar 20, 2025 at 06:23:51AM -0700, Grégory Mantelet wrote: gmantele left a comment (ivoa-std/ivoatex#153) You seem to have a perfectly clear idea on how to do that. So, if ok with you, I let you do it.

Well... "perfectly clear" is a very strong statement given it's involving the github CI. So... could you review PR #154? Thanks!

[msdemlei]

@mcdittmar, could you see if things are fixed now?

You'll have to update both ivoatex and the ci, so you would run

make update
make update-ci

(don't worry about the question whether to remove the CI, just say yes).

Oh, and if it works, would you close this bug?