seccomp mess, continued, status update

Hal Murray hmurray at megapathdsl.net
Mon Feb 24 06:34:22 UTC 2020


> Wouldn't it be simpler to ude a base image in the CI that isn't buggy? 

Maybe.  I don't know that area.  If that is the only place we test seccomp, 
then yes, we should switch to Fedora or Debian.  If that is testing if we can 
build on Alpine, then it has found a bug but the bug is in Alpine rather than 
our code.  We could maybe backup to an older version of Alpine, but I think it 
would be better to test their latest/current version and document that seccomp 
doesn't work there.

I'm working on a hack fix.

For the record, here is diff between Fedora and Alpine:

[murray at hgm ~]$ diff seccomp-syscalls.h-fedora seccomp-syscalls.h-alpine
275d274
< #define __PNR_ppoll                           -10241
1361,1366d1359
< #endif
< 
< #ifdef __NR_ppoll
< #define __SNR_ppoll                   __NR_ppoll
< #else
< #define __SNR_ppoll                   __PNR_ppoll
[murray at hgm ~]$ 


-- 
These are my opinions.  I hate spam.





More information about the devel mailing list