test results web page

[Joe Buck]

> I don't know how you store stuff internally, but it occurs to me
> that if you get lots of testlogs, your disk is going to get full
> quickly :-) :-)

If I understand right, he's storing the output of Alexandre's script
(e.g. only the FAILs and XPASSes), not the whole summary.

While I know it's tricky, I'd like to be able to do queries on individual
failures: e.g. show all tests that FAIL on sparc-* that do not FAIL on
ix86-* and the like.  This might mean having a table with rows organized
as follows:

testID, version, platform, test-name, test-status

with one line per individual test, where test-status would be either FAIL
or XPASS (there would be no row for passing tests).  I believe that this
table can now be queried to get the information.

testID would be an integer that is assigned to each submitted test,
and the idea is to use it for join operations -- have another table
with

testID, who, version, platform, testsuite-name, npass, nfail, nxpass, nxfail

something like

231, jbuck@synopsys.com, egcs-1.0.3a, sparc-sun-solaris2.5.1, gcc, 4321, 2, 0, 0

(make one line for each of gcc/g++/f77/etc so it would be possible to
record tests from people who don't build all front ends).