On Tue, Dec 15, 2015 at 11:07:24AM +0100, Holger Levsen wrote:
> Hi Mike,
> thanks for the bug report!
> On Dienstag, 15. Dezember 2015, Mike Hommey wrote:
> >    * What was the outcome of this action?
> > The output contained differences in the output for find due to
> > filesystem-dependent behavior wrt file ordering in readdir(3).
> > 
> >    * What outcome did you expect instead?
> > The sorted list should be compared instead.
> I think it would be fine if there was an option to force such a behaviour, 
> but 
> I think the default should be not to do that. diffoscope should show (all) 
> differences, not hide some, to make some others more visibe… at least not by 
> default.
> IOW: if your "container" (the filesystem) is underministic, diffoscope should 
> not try to hide this (by default).

The tricky thing is that it's not necessarily desirable that all
containers be treated equal during one run.

To give an actual example, I'm comparing Firefox builds that I get out
of Mozilla automation. They come in the form of a tar.bz2. Unpacked or
not, I don't actually care about the file order in them. However, I *do*
care about the file order in omni.ja and *.xpi files that are inside of

So a global option that treats all containers equal would not work
for my use-case.

That being said, I would care less about the `find {} -printf %P` diff
if there were other indicators that files are missing in one of the
containers. Currently, that information is essentially hidden by the
fact that the containers are not ordered the same.


Reproducible-builds mailing list

Reply via email to