----- Original Message -----
From: "Jerry van Dijk" <jvandyk@...>
Sent: Tuesday, 3 September 2002 01:48
Subject: Re: [MinGW-dvlpr] gnat 3.3 test results
> Executive summary:
> TESTS: 2397
> PASSED: 2260
> FAILED: 105
> NOT-APPLICABLE: 32
> For a first testrun ever on Windows this is actually pretty good. See
> test report for more details. More testruns and analysis are needed
> conclusions, but I would be confortable with a gnat stand-alone 3.3
> experimental release.
> I do note that executables have become rather bulky compared to 2.8.1.
> compiled testsuite took over 4Gb, which is at least 1.5 Gb more.
One reason for the increase in size is that during experimental phase
of gcc development cycle, extra RTL and runtime tree checking code is
enabled by default. Once it gets to the prerelease stage, this code is
disabled. (BTW, g-spipat.adb compiles fine at -O2 with the checking
disabled but when enabled, verify_flow_info finds problems with
inlining -- hence the addition of a special rule for g-spipat.o).