Hello,
Tanguy Le Carrour <
address@hidden> ezt írta (időpont: 2019. nov. 13., Sze 9:03):
Hi Marius, hi Gábor!
Le 11/04, Gábor Boskovits a écrit :
> Tanguy Le Carrour <address@hidden> ezt írta (időpont: 2019. nov. 4., H, 11:50):
> > Le 11/04, Gábor Boskovits a écrit :
> > > Tanguy Le Carrour <address@hidden> ezt írta (időpont: 2019. nov. 4.,
> > > > `./pre-inst-env guix build --rounds=2 python@3.8` always fails with:
> > > > """
> > > > output ‘/gnu/store/…-python-3.8.0’ of ‘/gnu/store/…-python-3.8.0.drv’
> > > > differs from previous round
> > > > """
> > […]
> > First time I hear of diffoscope! I had a look at the output (html
> > attached), but couldn't make sense of it!
> Nice, one thing that looks problematic is that there seems to be a mismatch
> in the hash.
> At one line you can see the shabang modified to a different path...
> I don't yet know the reason of that, but it quite suspicious...
> One thing, when dealing with things like like this, one should first fix
> the readable differences,
> and only then dig deeper, as fixing those often cascade, and fix more than
> anticipated.
Sorry it took me so long, but I finally had some time yesterday evening to go
through the report!
I found 4 types of errors:
1. `stat` reporting a different "Birth" for folders, which I assume are
perfectly normal.
This is fine, you can use the option to ignore directory metadata from diffoscope output, so that this won't pollute the output.
2. different paths for `python3.8` in the shebang of some scripts.
3. different paths in prefix for `python3.8` (some origin as above).
3 and 2 should be investigated. Do you get these diffs even with grafts disabled?
4. everything else is "Max report size reached"!
Should not be a problem, it shows the diff is too big. You have diffoscope options to increase that, but as problems are fixed the output should return to the normal range.
I don't really know where to go from here? Am I supposed to investigate
errors of type 4?!
Regards,
--
Tanguy
Best regards,
g_bor