Bug 239323

Summary: math/scilab "runaway process" during config
Product: Ports & Packages Reporter: dgilbert
Component: Individual Port(s)Assignee: Max Brazhnikov <makc>
Status: New ---    
Severity: Affects Some People CC: w.schwarzenfeld
Priority: --- Flags: bugzilla: maintainer-feedback? (makc)
Version: Latest   
Hardware: Any   
OS: Any   
Attachments:
Description Flags
12.0 poudriere build log
none
the options for math/scilab that lead to failure none

Description dgilbert 2019-07-19 23:24:42 UTC
Created attachment 205911 [details]
12.0 poudriere build log

In many of my recent poudriere builds, scilab has failed for a "runaway process" violation.  The process in the configuration portion of the build that is causing this is testing arpack-ng compatibility... and runs with 8 or more threads (only one busy).  I'm pretty sure I'm not geared to analyze this.

I've attached the poudriere biuld log for 12.0, but I've encountered this on 11.2 and now recently 11.3.

I've tried quadrupling my timeout... it doesn't give any evidence that it's working.
Comment 1 Max Brazhnikov freebsd_committer 2019-08-01 18:09:42 UTC
(In reply to dgilbert from comment #0)
> Created attachment 205911 [details]
> 12.0 poudriere build log
> 
> In many of my recent poudriere builds, scilab has failed for a "runaway
> process" violation.  The process in the configuration portion of the build
> that is causing this is testing arpack-ng compatibility... and runs with 8
> or more threads (only one busy).  I'm pretty sure I'm not geared to analyze
> this.

Apparently test program loops forever. I see you build scilab with OPENBLAS option. Do you have the same problem with default set?
Comment 2 Walter Schwarzenfeld freebsd_triage 2019-08-14 22:49:30 UTC
Should be solved with ports r508978.
Comment 3 dgilbert 2019-08-17 20:33:22 UTC
I think I need to clarify?  Are you asking me to rebuild scilab with all default options or rebuild it with whichever blas was the default?  Which blas is the default again?
Comment 4 dgilbert 2019-08-17 23:40:12 UTC
So... my result: it builds with entirely default options (but I assume you knew that) but it fails with my options.  I'm attaching my options.
Comment 5 dgilbert 2019-08-17 23:41:02 UTC
Created attachment 206648 [details]
the options for math/scilab that lead to failure
Comment 6 dgilbert 2019-08-18 19:46:51 UTC
after testing and rebuilding a number of times, I've narrowed down the cause of the failure.  the 64 bit integer option of openblas causes the problem.