Bronek Kozicki wrote:
Peter Dimov <pdimov@mmltd.net> wrote:
We don't seem to be running any tests on Comeau C++, why? Not using the "reference compiler" for our testing doesn't make much sense.
I can run, and in fact I was running regression tests on como 4.3.3 some time before release 1.32 . There are however some issues with it: 1 como in "strict" mode does not support platform specific headers (at least on Windows, and this is my testing environment) 2 como does not support dynamic libraries at all; this is going to be improved with next version (hopefully to be released soon) 3 libcomo (Comeau implementation of the C++ standard library) is far from perfect, but is improving 4 como, at least with my backend (MSVC) has worst compilation times ever seen; in fact, running regression tests with como was real PITA 5 there does not seem to be big interest on developers side to take como results into account (but that's my totaly subjective feeling - I know that there are developers who care about como results, and Greg Comeau seemed to be also interested in good compatiblity with boost)
I understand your concerns, but it seems unfair not to test anything because some libraries have problems on como. Is it possible to setup a como test suite that only tests libraries that aren't affected by (1) or (2)? This would presumably address (4) as well. As for (5), I'm very interested in seeing the test results of running a strict EDG-based compiler. It is true that few of us use como as our primary compiler, but it helps people write correct code, instead of code that just happens to compile on our test matrix. The latest Metrowerks and g++ seem promising, but I'm not sure whether they can displace EDG as a test platform yet when strict conformance is desired.