Bug 689167

Summary:	Diff'ing compiler output as part of regular regression testing.
Product:	Ghostscript	Reporter:	Timothy Osborn <tim.osborn>
Component:	Regression	Assignee:	Robin Watts <robin.watts>
Status:	RESOLVED FIXED
Severity:	enhancement	CC:	christinedelight.top85, tim.osborn
Priority:	P4
Version:	master
Hardware:	PC
OS:	Linux
Customer:		Word Size:	---

Description Timothy Osborn 2007-04-11 11:36:03 UTC

It would be beneficial to compare compiler output as part of our regression
suite. At least this would let us see the differences in warnings so that we can
get a handle on warning growth. Also... most likely (but not always, of course)
the new warnings are the significant ones that you want to see, but are lost in
the mix.

What brings this up is there were some significant warnings missed in the 7819
build. This sort of mechanism would help in these situations as finding the bugs
caused by some of these issues is time-consuming work.

Note: Peter made the following comment when I suggested this idea:

"I think that is a fine idea.  Note that any change in a source file will
cause the line numbers of warnings to change, so the line numbers should be
removed before running 'diff' (easy to do with a script)."

Comment 1 Ray Johnston 2010-04-25 22:36:50 UTC

I think that bmpcmp mostly accomplishes this.

If so, please close this bug.

Comment 2 Robin Watts 2010-06-01 11:36:38 UTC

I read this as being about compiler warnings etc, rather than ghostscript output. As such, I can't see how bmpcmp applies here?

Marcos has a script to produce a list of new warnings etc. Maybe all that's needed here is for us to use that to count warnings on each SVN/localcluster build, and display a "delta" figure in the report?

"This change caused 23 new warnings: Here is a list of some of them:" etc

Comment 3 Robin Watts 2011-07-20 14:34:55 UTC

I believe that this bug refers to keeping track of warnings etc. As such Marcos already does this. So I'm closing, (prompted by Shailesh).

Thanks.