On 2024/08/07 16:08:02 +0200, Stefan Sperling <stsp@stsp.name> wrote:
> On Wed, Aug 07, 2024 at 03:01:03PM +0200, Omar Polo wrote:
> > > +	/*
> > > +	 * SHA2 repositories cannot be used with gotd until Git protocol v2
> > > +	 * support is added. Reject them at startup for now.
> > > +	 */
> > > +	TAILQ_FOREACH(repo, &gotd.repos, entry) {
> > > +		struct got_repository *r;
> > > +
> > > +		error = got_repo_open(&r, repo->path, NULL, NULL);
> > > +		if (error)
> > > +			fatalx("%s: %s", repo->path, error->msg);
> > > +
> > > +		if (got_repo_get_object_format(r) != GOT_HASH_SHA1) {
> > > +			error = got_error_msg(GOT_ERR_NOT_IMPL,
> > > +			    "sha256 object IDs unsupported in network "
> > > +			    "protocol");
> > > +			fatalx("%s: %s", repo->path, error->msg);
> > > +		}
> > > +
> > > +		got_repo_close(r);
> > > +	}
> > 
> > didn't we also allowed repositories to be listed but not yet created?
> > or maybe i'm remembering wrong.
> 
> In theory, yes, the intention was to do this. Thanks for reminding me.
> However, it turns out I already broke this functionality back in commit
>   4b3827cd43394b89d2af822dcd1d9a9179c1ee10
>   "make gitwrapper ignore 'permission denied' for repository paths"

ah, ops, sorry for forgetting this while reviewing that diff.  maybe we
need to add a test :P

> Since then parse.y runs yyerror() when a repository does not exist,
> causing gotd to exit on startup. Which can be fixed separately.

sure

> Here is a new version of the proposed diff which ignores ENOENT.

ok op@, thank you!