[Fuego] A few fixes and a bug report

Tim.Bird at sony.com Tim.Bird at sony.com
Fri Oct 5 01:39:26 UTC 2018



> -----Original Message-----
> From: Daniel Sangorrin
> 
> Hi Tim,
> 
> I did a few modifications and found a regression.
> 
> [PATCH 1/3] dynvars: add support for simple variable assignments
>   -> this implements your proposal about using simple variable
>      assignments for dynamic variables
> 
> [PATCH 2/3] run-test: use makedirs instead of mkdir
>   -> I found cases where the "builds" folder had not been created
>      so use os.makedirs instead to get an "mkdir -p" like functionality.
> 
> [PATCH 3/3] ovgen: remove test spec strip and generation code
>   -> This removes some legacy code. If you want to keep the legacy code
>      I am fine, you could leave a "this is deprecated" kind of message
> 
> About the regression. I already mentioned it once, but it seems that there
> is a regression or at least some problem with the criteria handling code.
> Benchmark.Dhrystone fails with these messages:
>    Error: criteria missing reference value - returning SKIP
>    ERROR: results did not satisfy the threshold

OK - I did a quick peak at the patches, and they look OK.  I won't
have time to apply and test until tomorrow.  I'm in the middle of a
big sweep through the ftc and core code changing logging, and introducing
a new phase 'snapshot'.  Details to follow.  One of the reasons for changing
the output is that it was hiding regressions like the one you found.  I
suspected that something in the criteria code broke - but I'm 
not sure when.  I'll try to find some time to look at it tomorrow.
 -- Tim



More information about the Fuego mailing list