Hello,
I'm currently playing with smoking different gcc versions.
My question: what would the best approach for this be?
I ran a smoke using a configuration file with:
=
-Dcc='ccache gcc'
-Dcc='ccache /tmp/GCC/D/bin/gcc'
The output in mktest.rpt:
darkstar: Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz (GenuineIntel
2400MHz) (i686/4 cpu)
on linux - 2.6.22-smp [slackware]
using ccache gcc version 4.1.2
smoketime 26 minutes 14 seconds (average 13 minutes 7 seconds)
...
33662 Configuration (common) none
----------- ---------------------------------------------------------
O O - - -Dcc='ccache gcc'
O O - - -Dcc='ccache /tmp/GCC/D/bin/gcc'
| | | +----- PERLIO = perlio -DDEBUGGING
| | +------- PERLIO = stdio -DDEBUGGING
| +--------- PERLIO = perlio
+----------- PERLIO = stdio
The output in mktest.out:
...
Configuration: -Dusedevel -Dcc='ccache gcc'
------------------------------------------------------------------------------
Compiler info: ccache gcc version 4.1.2
...
Configuration: -Dusedevel -Dcc='ccache /tmp/GCC/D/bin/gcc'
------------------------------------------------------------------------------
Compiler info: ccache /tmp/GCC/D/bin/gcc version 4.2.3
...
The header in mktest.rpt only lists 4.1.2 and not 4.2.3.
This brings me to my question:
What would the prefered solution be?
a) running multiple smokes - one for each versions of gcc?
b) A patch for the report builder to output each gcc versions? (which
would also mean patching the summary builder and so on)
(My current testing idea is to install each version of gcc and smoke it.
When that is up and running I'm also thinking of installing the gcc
snapsnots and testing those (first the gcc test, then a perl smoke). The
downside of this last one is - of course - that some builds will fail due
to unresolved/new bugs in gcc.)
Kind regards,
Bram
Thread Next