[RFC/RFH] IMA testsuite

[Zack Weinberg]

Andrew Pinski <pinskia@physics.uc.edu> writes:

> Since we don't have a way to test that IMA is working I decided to try
> add a way.  This patch adds a simple IMA testsuite (well only one test
> right now).  But I need help adding more complex ones like ones where
> we can do dg-final and dg-error.
>
> Also what do people think of this patch and what can people do to help,
> I don't know TCL or expect that much?

This is a good start.  I would suggest three things:

1) Use the higher-level primitives in dg.exp (part of dejagnu, not in
   lib/).  This is how you get dg-* markers working.

2) Implement iteration over optimization levels.

3) We want to test all possible permutations of all possible clusters
   of files.  For a two-file test case, that's

   - [A.c] [B.c]
   - [A.c B.c]
   - [B.c A.c]

   and for a three-file test case,

   - [A.c] [B.c] [C.c]
   - [A.c B.c] [C.c]
   - [B.c A.c] [C.c]
   - [A.c C.c] [B.c]
   - [C.c A.c] [B.c]
   - [B.c C.c] [A.c]
   - [C.c B.c] [A.c]
   - [A.c B.c C.c]
   - [B.c A.c C.c]
   - [B.c C.c A.c]
   - [C.c B.c A.c]
   - [C.c A.c B.c]
   - [A.c C.c B.c]

   where square brackets surround groups of files to be passed to the
   compiler simultaneously; order matters within square brackets but
   not between square brackets.

   Yes, this does combinatorially explode, but I doubt there will be
   bugs that only show up with 4 input files, so we should be able to
   get away with handling only 2 and 3 files, and the sets aren't that
   big for those.

zw