Re: "make check" failure in public-submodule-commit

[Stefano Lattarini]

On 11/20/2012 12:13 PM, Théophile Ranquet wrote:
> Hi Stefano, hello Gnulibers,
> 
> 2012/11/19 Stefano Lattarini <address@hidden>:
>> $ git submodule --quiet foreach test '$(git rev-parse $sha1)'  = '$(git 
>> merge-base origin $sha1)'
> 
> The quoting here seems a bit lacking. This is an issue with gnulib,
> please see the attached patch for a fix. Note that the LRDE has filed
> a global disclaimer for its staff for both Bison and Gnulib, so my
> contributions can applied.
>
>> [...]
>> + eval test '$(git rev-parse $sha1)' = '$(git merge-base origin $sha1)'
>> +++ git rev-parse daf7f8c02242c535d596231e2f655109b97fa2bc
>> +++ git merge-base origin daf7f8c02242c535d596231e2f655109b97fa2bc
>> fatal: Not a valid object name origin
> 
> This is something else, and I haven't been able to reproduce this bug.
> Have you bootstrapped or run 'git fetch' in bison/gnulib/ recently?
>
Bootstrapped, yes, after a "git clean -fdx" (I did to verify the failure
I was experiencing was not due to some inconsistency in the build
directory).  I never run 'git fetch' in the 'gnulib' subdirectory though;
I expect the bootstrap script to do that for me automatically (as is done
in other projects, I think).  Anyway, this evening I'll try again with a
freshly-cloned bison, to see whether the error still persist.

>> ++ test daf7f8c02242c535d596231e2f655109b97fa2bc =
>> /usr/local/bleeding/libexec/git-core/git-submodule: line 445: test: 
>> daf7f8c02242c535d596231e2f655109b97fa2bc: unary operator expected
> 
> This error is gratuitous, and related to the quoting issues mentioned above.
> 
And is fixed by your patch :-)

Thanks,
  Stefano