dropping bup from s390

[Jon Dowland]

Hi,

My package 'bup' has not made it to testing due to FTBFS on s390: #664185

I've asked upstream about it and they don't have the time/knowledge/inclination
to fix it, or support s390 in general. To be honest I don't either.

Therefore, I plan to remove s390 as a supported architecture.

Please shout if you would prefer me not to / can help fix that bug in the next
few weeks before freeze.


Thanks

[Bastian Blank]

Post by Jon Dowland
My package 'bup' has not made it to testing due to FTBFS on s390: #664185

The test suite allocates too much memory. Nothing mythical about it.
_Why_ it does this it up to you to find out.

Post by Jon Dowland
I've asked upstream about it and they don't have the time/knowledge/inclination
to fix it, or support s390 in general. To be honest I don't either.
Therefore, I plan to remove s390 as a supported architecture.

This is unfortunate. However for fundamental bugs (yes, using mmap to
map large files is a fundamental bug), this is no solution.

Bastian

[Jon Dowland]

The test tries to mmap a file with the size of slightly above 1GiB.
| create(): expected = 268435456
| create(): bits = 30
| create(): size = 1073741824
| ! tbloom.py:38 4 == 4

Yes.

The FTBFS can be worked around by skipping "make test" on s390. I'm
considering that.

You're right that the problem is not specific to s390 (it's just triggered a
lot earlier). I don't categorically know that bup itself (not just the tests)
gracefully handle failing mmaps. Upstream report that i386 can get into
problems with large backups. Whether that's RC is another question.

Upstream did map out a possible solution for gracefully handling failing
mmap within bup, but no code so far.

And waiting for 5 month is also no solution.

You are being unnecessarily antagonistic - please try to be constructive.