On Wed, Aug 28, 2019 at 12:51:07PM +0200, Karel Zak wrote:
> On Tue, Aug 27, 2019 at 02:26:08PM +0200, Patrick Steinhardt wrote:
> > Triggered by commit f612c4c67 (tests: fix --unbuffered mode with
> > ASAN, 2019-08-27), which says:
> > 
> >     Well, this patch sucks. It would be nice to have things in
> >     the way how it has been original expected by Patrick's patch,
> >     but ...
> > 
> > So this commit here effectively reverts it and instead tries to
> > improve the shortcomings of the original patch. First, it uses
> > env(1) to set ASAN_OPTIONS instead of directly adding it to the
> > args array to fix execution of "${args[@]}" "$@".
> > 
> > Second, it now supports both unbuffer(1) and stdbuf(1). The
> > latter uses LD_PRELOAD tricks, which doesn't play nicely with
> > ASAN, so it will not be used if ASAN has been requested. It's
> > still valuable to have support for both, as many more systems
> > will have stdbuf(1) from coreutils installed but not unbuffer(1)
> > from expect.
> 
> Applied, thanks.
> 
> Unfortunately, it seems result is fragile. It fails on travis
> https://travis-ci.org/karelzak/util-linux/jobs/577767328
> 
> The problem is the order of some output lines, for example:
> tests/output/rename/exit_codes
> 
> Maybe the best and most robust solution would be split stdout and
> stderr into two files for each test. Now we have $TS_OUTPUT, what about
> to replace it with $TS_STDOUT and $TS_STDERR? Something like:
> 
>  - $TS_CMD_MOUNT $DEVICE $TS_MOUNTPOINT >> $TS_OUTPUT 2>&1
>  + $TS_CMD_MOUNT $DEVICE $TS_MOUNTPOINT >> $TS_STDOUT 2>> $TS_STDERR
> 
> And also create tests/expected/<foo>/<bar>{.err} for tests where any
> output on stderr is expected, otherwise assume $TS_STDERR is empty.
> 
> This change will be pretty invasive, but than we can easily compare 
> outputs and keep track about what is where printed. 
> 
>     Karel

Sorry, I somehow didn't see this mail. I thought about splitting
it up too, but shied away as I expected it to be rather involved.
Anyway, I think in the long run it's definitely the best way to
tackle the issue at hand, so thanks for your effort!

I've just done a test run on musl libc, and everything except for
the known issue with POSIXLY_CORRECT getopt passes.

Patrick