On Tue, Feb 14, 2023 at 11:11:26PM +1100, Mark Jamsek wrote:
> On 23-02-14 12:47PM, Christian Weisgerber wrote:
> > Mark Jamsek:
> > 
> > > I made adjusting regress for the new fetch behaviour change more
> > > difficult than necessary because of a copypasta error where the reported
> > > failure returned a bogus test function. Silly error on my part but to
> > > figure out the cause of the failure I added $LINENO to the output such
> > 
> > Portability concern: Although LINENO is in POSIX, it's only covered
> > as part of an extension.  dash, the somewhat popular minimalist sh
> > on Linux, does not support it ... or at least older versions of
> > dash still used by stable versions of Debian don't support LINENO.
> 
> Thanks, naddy! I wasn't too sure of its portability impact so was hoping
> you or someone as knowledgeable would have some insight.
> 
> It's really not an important change; it just helped with a problem of my
> own making and I thought to share it in case it might have some broader
> utility that appeals to others. But our tests already do provide enough
> info to isolate any problems so I think it might be best to forget this,
> at least for now.
> 
> Thanks again :)

I agree that this would be nice to have, but portability is indeed
a concern for -portable. It would also be nice to have a solution
that only tweaks test_done in one place rather than having to pass
additional parameters. But I don't have no idea how that could be done.

Maybe it is time to rewrite the tests in Rust?