Git fetch failure in submodule if deeper submodule pointer updated

Git fetch failure in submodule if deeper submodule pointer updated

[Yngve N. Pettersen]

[-- Attachment #1: Type: text/plain, Size: 1059 bytes --]

Hello all,

Recently we encountered a git fetch issue similar to the one I have  
reported earlier, <https://marc.info/?l=git&m=158979416620251&w=2>, which  
AFAICT has not yet been fixed.

In this case we had checked out a submodule, but not the submodule(s)  
below it.

The full submodule chain was like this:

   top->middle->bottom

The actual checkout was

   top->middle

Because only "middle" was needed for the cron job script used to push  
updates to "top" and "middle", "bottom" was never checked out (and it  
should not be necessary to do so, either).

When the pointer to "bottom" was recently updated in "middle", the cron  
job failed, because Git "could not access submodule bottom".

As I said in my earlier report, this kind of issues should, at most, only  
trigger a warning, not a fatal error.


The Git version on the system is Git v2.25.1 (Ubuntu 20.4)

This problem is not occuring in Git v2.17 on Windows.

Attached is a zipfile with a script that reproduces the problem.



-- 
Sincerely,
Yngve N. Pettersen
Vivaldi Technologies AS

[-- Attachment #2: generate_test.zip --]
[-- Type: application/zip, Size: 548 bytes --]

Re: Git fetch failure in submodule if deeper submodule pointer updated

[Philippe Blain]

Hello Yngve,

Le 2020-12-19 à 09:37, Yngve N. Pettersen a écrit :
> Hello all,
> 
> Recently we encountered a git fetch issue similar to the one I have reported earlier, <https://marc.info/?l=git&m=158979416620251&w=2>, which AFAICT has not yet been fixed.

Regarding your earlier report, I don't think it has been fixed either.
The fact is that I don't think any full-time contributor to Git works
on submodules much these days, so issues like this one, that need design-level
discussions to be fixed, often take more time to get fixed, if they ever are.
Now on to your new report:

> 
> In this case we had checked out a submodule, but not the submodule(s) below it.
> 
> The full submodule chain was like this:
> 
>    top->middle->bottom
> 
> The actual checkout was
> 
>    top->middle
> 
> Because only "middle" was needed for the cron job script used to push updates to "top" and "middle", "bottom" was never checked out (and it should not be necessary to do so, either).
> 
> When the pointer to "bottom" was recently updated in "middle", the cron job failed, because Git "could not access submodule bottom".
> 
> As I said in my earlier report, this kind of issues should, at most, only trigger a warning, not a fatal error.
> 
> 
> The Git version on the system is Git v2.25.1 (Ubuntu 20.4)
> 
> This problem is not occuring in Git v2.17 on Windows.
> 
> Attached is a zipfile with a script that reproduces the problem.

Thanks for the report and the reproducer. This bug was in fact reported and fixed
earlier this month [1] (full threads: [2], [3]). The patch series is currently cooking
in the 'next' branch [4]. You can check out just the fix from [5]. I don't think
it's going to make it to Git 2.30, but it should be in 2.31.

Cheers,

Philippe.

[1] https://lore.kernel.org/git/20201209105844.7019-1-peter.kaestle@nokia.com/
[2] https://lore.kernel.org/git/CAN0XMOLiS_8JZKF_wW70BvRRxkDHyUoa=Z3ODtB_Bd6f5Y=7JQ@mail.gmail.com/t/#u
[3] https://lore.kernel.org/git/1604413399-63090-1-git-send-email-peter.kaestle@nokia.com/t/#u
[4] https://github.com/git/git/blob/730f2a8a60960c30a79e00ebe034836f60befbf0/whats-cooking.txt#L765-L771
[5] https://github.com/gitster/git/tree/pk/subsub-fetch-fix-take-2