Control: tags -1 fixed pending Greetings,
I took some time to reset my machine to the new default pid_max,
and with a build of t-coffee with an increased MAX_N_PID to
4194304. This approach seems to work, because the program
properly outputs its version without crash or memory saturation:
$ bash -c 'echo $$' && t_coffee -version && bash -c 'echo $$'
1012124
PROGRAM: T-COFFEE Version_13.41.0.28bdc39 (2019-11-30 10:21:53 -
Revision 5d5a1c1 - Build 465)
1012132
The run-unit-test script works under high PID number conditions
as well. My quick understanding of the code is that this number
is used for building a sparse table of processes, which would
grow from 2M to 33M: it is the only drawback I can think of from
that approach, which might remain reasonable (although it might
be a wee bit more a concern on 32 bits CPU).
The updated package is pushed on Salsa and ready for review:
https://salsa.debian.org/med-team/t-coffee
I could not pinpoint the exact moment the default pid_max change
occurred in Debian Bullseye, but the new value of 4194304 is
there and it looks like it will become common in the next few
generations of systems.
That value seems to be a structural limit within Linux on 64bits
machines, so don't seem to be bound to grow further anytime
soon. From the patch in https://lkml.org/lkml/2020/4/8/344:
| + * Linux limits the maximum number of tasks to PID_MAX_LIMIT, which is
currently
| + * 0x400000 (and can't easily be raised in the future beyond FUTEX_TID_MASK).
Kind Regards,
--
Étienne Mollier <etienne.moll...@mailoo.org>
Fingerprint: 5ab1 4edf 63bb ccff 8b54 2fa9 59da 56fe fff3 882d
Help find cures against the Covid-19 ! Give CPU cycles:
* Rosetta@home: https://boinc.bakerlab.org/rosetta/
* Folding@home: https://foldingathome.org/
signature.asc
Description: PGP signature
