[BUG] oiiotool not returning correct error code for wildcard + missing file

[lgritz]

Paraphrased from a discussion on the Slack where a user said:

Today I stumbled on an interesting behaviour in oiiotool that shadowed another (internal) bug and got my head scratching for a bit.
Image you’re doing a simple operation, but none of the input files exist. oiiotool will nicely report that as an issue:

$ oiiotool --runstats non-existent.exr non-existent.exr  --over -o test.exr
oiiotool ERROR: read : File does not exist: "non-existent.exr"

$ echo $?
1

However, if you use the frame syntax to refer to non existent frames, it will not flag any issue, and exit with 0:

$ oiiotool --runstats  non-existent.%04d.exr non-existent.exr --over -o test.exr
oiiotool WARNING : oiiotool produced no output. Did you forget -o?

$ echo $?
0

I tested this is with

$ oiiotool --version
2.5.14.0

Not a biggie, but I wonder if the single-file check that happens in the first case could be expanded to work with the supported syntaxes for frames too?

[lgritz]

I'd need to look, but what it feels like to me is that maybe the widcard loop over frames is somehow saying it succeeded even if one of the individual loop iterations fails? (Or maybe even if all of them fail?)

[lgritz]

OK, got a chance to look at this. I'm not sure it's really a bug.

When the command line contains any names that use frame wildcards, it finds all the files that match the pattern, then runs the set of commands once for each matching file.

$ oiiotool --runstats  non-existent.%04d.exr non-existent.exr --over -o test.exr

If non-existant.0002.exr had existed and been found then it would have tried to run the commands for frame 0002, and that would have produced an error because the second file, non-existent.exr, did not exist. But there would be no error for frame 0001 or frame 0003 because there IS NO frame 1 or 3, so the loop commands aren't executed at all.

Stated plainly: The pattern is used to establish the set of frame numbers (by which files match the pattern and what there numbers in those files are), then the full set of commands is run for each of those frame numbers (with the patterned files being replaced by the right number), and if any of those commands has an error when executed (like not being able to read a file), then you see an error.

So because no frame numbers were found to match your pattern, the frame number list was empty and there was never an attempt to run your commands, so it never tried to read the non-wildcard file.

I think this is the right behavior, because it's ok to have a frame wildcard not match every number.

But... I do see your point. Maybe this boils down to an issue of whether it should be considered an error if there are NO matching frame numbers at all for your wildcard? So the error message in that case wouldn't be about non-existant.exr not being found, it would be more like "No frames were found to match the wildcard pattern". Does that sound like the right solution to you?

[lgritz]

Hang on, I think I see a simple solution that will improve this situation...