Bug#945041: openfoam: severe build time regression

[Adrian Bunk]

On Thu, Dec 19, 2019 at 12:31:09AM -0600, Kurt Kremitzki wrote:
> On Thursday, December 19, 2019 12:11:33 AM CST Adrian Bunk wrote:
> > On Mon, Nov 18, 2019 at 10:17:10PM +0100, Julien Cristau wrote:
> > > Source: openfoam
> > > Version: 1906+dfsg1-1
> > > Severity: serious
> > > 
> > > Hi,
> > > 
> > > Looking at https://buildd.debian.org/status/logs.php?pkg=openfoam the
> > > build time for your package increased unreasonably in this version.
> > >...
> > 
> > Something seems to be going wrong around parallel building,
> > when trying locally only one core gets used now.
> 
> Indeed, although the build logs state e.g. 'Compiling enabled on 4 cores', 
> only 1 core is ever used. However exporting the variables defined in debian/
> rules and following the same build steps in a shell results in a build which 
> does actually use the number of cores it reports, so I wasn't sure where to 
> proceed with troubleshooting it beyond that. I mentioned the issue to my 
> upstream contact as well who wasn't sure either given that it behaves as 
> expected in a shell but not in the build environment.

wmake/wmake:
...
if [ -n "$WM_NCOMPPROCS" ]
then
    parOpt="-j $WM_NCOMPPROCS"

    if [ "$WM_NCOMPPROCS" -gt 1 -a -z "$MAKEFLAGS" ]
    then
        make="$make --no-print-directory $parOpt"
    fi
fi
...


The problem is that MAKEFLAGS is set to "w",
as fix/workaround add in debian/rules:
  unexport MAKEFLAGS

cu
Adrian