First attempt at new Sunday test project

First attempt at new Sunday test project

[Elena Zannoni]

Hi, I used Michael Chastain's scripts to generate some test results.
It is all still very rough, because it took me a long time to get the
builds done. As you can see, almost 2 weeks. I think I can streamline
the process a bit better now. I did it in small increments, a few
minutes each day on a very slow machine. The idea is that this can
help GDB6 getting out the door faster.


Anyway, these preliminary ballpark results are in:

http://people.redhat.com/~ezannoni/REPORT-06162003/index.html

(note: there is no Analysis.txt file, I haven't analyzed anything).


Let me know if I should change the combination of things I am testing.
For instance I have both gcc3.3 and gcc-3_3-branch. Also, I have
binutils-2_14-branch, but this can be changed to use the released
binutils-2.14. I wonder if I missed some combinations.

Michael, if you can eyball something wrong in the setup, etc, please
let me know. 

elena

Re: First attempt at new Sunday test project

[Michael Elizabeth Chastain]

Aw, cool!!

If you have any feedback for me on using the scripts I would welcome it,
as you are the second user (first one who is not me).  I'm sure there is
a lot of confusing stuff and a lot of under-documented stuff.

You have to run gdb 5.3 so that we can look for regressions from
gdb 5.3 to gdb HEAD.  The "compare by gdb" table is what I use to
see if gdb HEAD is releasable.

It would be helpful to add gcc 2.95.3.

Testing with a lot of different binutils is a low-yield use of machine
resources.  That is, I hardly ever see differences based on binutils, and it
adds a lot of time.  So if you are low on CPU time then you could cut the
binutils down to just the vendor version.

Similarly you can cut down on the gcc's if you need to save CPU time
or disk space.

So I would recommend:

  gcc 2.95.3
  gcc 3.3
  gcc HEAD

  binutils 2.14

  gdb 5.3
  gdb HEAD
  gdb gdb_whatever_branch

  -gdwarf-2
  -gstabs+

That will get you about 90% of the good coverage for seeing if
gdb has problems.

But this leaves out the vendor gcc + vendor binutils combination which
might be important because of NPTL.  You might want to add those.

gcc HEAD with -gstabs+ has a serious issue.  I filed a gcc PR about it.
Don't remember the number off hand.  Looks like it's not fixed yet.

Michael C