Bug#1030325: lintian: archive-liberty-mismatch (in section for firmware-nvidia-gsp) non-free-firmware vs non-free
On 05/02/2023 08.47, Axel Beckert wrote:
Thanks for the lengthy analysis!
IMO only these cases should appear in the archive:
source:$area binary:$area # same $area, mandated by policy
source:main binary:contrib # iirc there is a policy exception for that
source:non-free binary:non-free-firmware # recently added, may need to
be addressed in the policy, together with non-free-firmware
any other source:$area1 binary:$area2 combination (regardless of liberty
order) should be prohibited (i.e. E: archive-liberty-mismatch)
There is one special case: for source:$area1 binary:$area2 mixed area
exceptions there must be at least one other binary2:area1 package
(i.e. source must build at least one package in its area).
Do the existing test cases cover all these cases?
There should probably be
for sarea in $areas
for barea in $areas
try Source:spkg #Section:
Package:pkg1 Section:foo/$sarea
Package:pkg2 Section:foo/$barea
try Source:spkg #Section:
Package:pkg2 Section:foo/$barea
Package:pkg1 Section:foo/$sarea
try Source:spkg #Section:
Package:pkg2 Section:foo/$barea
try Source:spkg Section:foo/$sarea
Package:pkg1 #Section:foo/$sarea
Package:pkg2 Section:foo/$barea
try Source:spkg Section:foo/$sarea
Package:pkg2 Section:foo/$barea
That should generate tests for all (im-)possible combinations.
The first three are for guessing the area from the first binary package
if the Source stanza has no section.
Will apply that one with some fine-tuning wrt. to line length, etc.
And then release Lintian 2.116.3.
Please include the dkms patch (see other bug), too.
On 05/02/2023 06.47, Axel Beckert wrote:
> Unfortunately firmware-nvidia-gsp/nvidia-graphics-drivers is rather
> large and I ran out of disk space when first unpacking it in my home
Then you should try src:nvidia-cuda-toolkit ;-) Lintian needs 3-4 hours
and about 50GB in /tmp to process a binary build. On nvme.
I'll probably come back with a bug report about lintian crashing (after
a few hours) when processing cuda 12 (but first need to check with the
latest lintian version again and also check that there was enough free
space in /tmp)
Andreas
Reply to: