RE: [Breakage] Git v2.21.0-rc0 - t5318 (NonStop)

[Randall S. Becker]

On February 8, 2019 13:03, Jeff King wrote:
> To: Randall S. Becker <rsbecker@xxxxxxxxxxxxx>
> Cc: 'Junio C Hamano' <gitster@xxxxxxxxx>; git@xxxxxxxxxxxxxxx; 'Linux
> Kernel' <linux-kernel@xxxxxxxxxxxxxxx>; git-packagers@xxxxxxxxxxxxxxxx
> Subject: Re: [Breakage] Git v2.21.0-rc0 - t5318 (NonStop)
> 
> On Fri, Feb 08, 2019 at 12:49:59PM -0500, Randall S. Becker wrote:
> 
> > > We did discuss this at the time of the patch, but it seems we
> > > already use /dev/zero in a bunch of places:
> > >
> > >
> > > https://public-inbox.org/git/xmqqbm57rkg5.fsf@xxxxxxxxxxxxxxxxxxxxxx
> > > com/
> > >
> > > Were you just skipping the other tests before?
> >
> > I did not catch the implications of the review at the time - my bad. We
> were not intentionally skipping the tests. It looks like some are automatically
> skipped. t4153 automatically skips (missing TTY), and t5562 fails also but for
> a different reason (hang - we don't have apache2 to serve up http content).
> >
> > Would you object to something like this:
> >
> > if [ ! -e /dev/zero ]; then
> > 	# use shred or some other mechanism (still trying to figure out a
> > solution) else
> > 	# existing dd
> > fi
> 
> That's fine, as long as it's wrapped up in a function in order to keep the tests
> readable.
> 
> Though I suspect we may be able to just find a solution that works
> everywhere, without having two different implementations. If we know we
> need $count bytes for dd, we could probably just generate a file with that
> many NULs in it.

For this, we could use truncate -s count file instead of dd to get a fixed size file of nulls. This would remove the need for /dev/zero in t5318 (the patch below probably will wrap badly in my mailer so I can submit a real patch separately.

@@ -383,7 +383,7 @@ corrupt_graph_and_verify() {
        cp $objdir/info/commit-graph commit-graph-backup &&
        printf "$data" | dd of="$objdir/info/commit-graph" bs=1 seek="$pos" conv=notrunc &&
        dd of="$objdir/info/commit-graph" bs=1 seek="$zero_pos" count=0 &&
-       dd if=/dev/zero of="$objdir/info/commit-graph" bs=1 seek="$zero_pos" count=$(($orig_size - $zero_pos)) &&
+       truncate -s $orig_size "$objdir/info/commit-graph" &&
        test_must_fail git commit-graph verify 2>test_err &&
        grep -v "^+" test_err >err &&
        test_i18ngrep "$grepstr" err

This passes on my platform.

> Other cases don't seem to actually care that they're getting NULs, and are
> just redirecting stdin from /dev/zero to get an infinite amount of input. They
> could probably use "yes" for that.

What about reading from /dev/null?

Regards,
Randall