Hi Adrian,

This catches me by surprise. I did rebuild the top1000 packages by
popcon and found exactly one FTBFS (which is fixed now). Given the low
number, we went ahead and then Santiago reached out regarding a few more
FTBFS and we figured the numbers were small enough.

Surprisingly, none of them were mpicc. Additionally note that many mpicc
users already do have patches and could be applying them.

Both uses you show are broken for cross compilation, so they need fixes
anyway. When using clang as compiler, the -target option should be
given.

In an ideal world, I'd agree, but compat levels are so rare and the
failure rate (that I saw) was so low that the compat level did not seem
like a suitable tool.

I searched current FTBFS bugs and found #1139535. Would you be able to
point to the multiple FTBFSes and also to the clang ones?

Helmut

Regarding FTBFS:
I've already fixed #1139537 #1139538 #1139540
The maintainer did a library transition for fixing #1139536

cu
Adrian

Adrian Bunk:

Hi,

Thanks for filing the bug and making us aware of the issue. Very
unfortunately that it caused you extra work unintentionally.

As I understand the situation, the unexpected breakage has been very
limited (<20) and has surfaced as FTBFS bugs (that is, loud breakage as
opposed to silent breakage). Additionally, every case known has already
been resolved.

All combined, makes it unclear to me that reverting is the best way
forward. Therefore, I have tagged it moreinfo and downgraded to
important for now. The latter to avoid annoying the RT team with a long
term unfixed RC bug.

Best regards,
Niels