Op een mooie herfstdag (Tuesday 12 November 2002 14:51), schreef Orton, Yves:
> Hi All. > > You may have noticed in the last smoke report that i sent out that the > version was marked 2.0 > > This is a provisional version number that I am using for a revamped (and > much more pure Win32 friendly) smoke test suite. To be honest, I am not too thrilled. I spend the better part of my free time since YAPC::Europe doing the same sort of thing. We should have co-ordinated our efforts better, but it looks like you've managed to get further down my TODO list than I have. Things being as they are, I'd be happy to hand over maintenance of Test::Smoke to you. It is a non-registered module so you can just upload it to CPAN. Abe > My intentions are > > 1. Convert the smoke test suite into modules. This is %50 complete. I am > currently calling them Test::Smoke::xxxx > 2. Reduce the scripts to be pure option handlers and then call into the > appropriate Test::Smoke::xxxx module. (50%) > 3. Provide support for nmake based builds. (Test-Smoke-1.16 is broken in > this regard.) (100% complete) > 4. Handle capturing of "dubious" test results. (100% complete) > 5. Remove all uses of global variables and enviornment variables (where > pos) from the code. This is to enable proper data flow analysis and to > facilitate future extensions to the scripts. (50%) > 6. Provide SMTP/Mime::Lite functionality to the smoking suite. (25%-- > required code framework changes have been made, but Mime::Lite support is > not implemented) > 7. Reduce and isolate the *nix biased parts of the testing suite to provide > for easier use on non *nix platforms. (50% complete) > 8. Provide propper batch file support for nmake/VDS6/VDS7. (100%) > 9. Provide better automatic extraction of compiler and OS version. (100%) > > In short im revamping the suite to be a bit more generic. > > A few thoughts: > > mktest.pl uses harness to extend the errors. From what I can see this is > superfluous. Am I correct? > > I have completely reworked mk_ovz.pl (i call my new version mk_rpt.pl and > Test::Smoke::Report) It uses different rules to parse mktest.out, > specifically focusing on the return table from harness instead of the more > difficult to parse error responses on single lines. So far this has been > successful. (At least in my build enviornment) > > I have completely reworked smokew32.pl to be more friendly towards non gcc > builds. Also added support for better option handling. The new version > however makes no attempt to work under non NT platforms. If a W9x user > starts to scream I will be happy to work with them to make a non NT > version, but I will not cripple the batch file to be W9x backwards > compatible. From what i can see there arent any W9x smoke tester anyway, > as the current smokew32.bat contains NT specific commands, so this shouldnt > be a problem. > > I am not sending out source code right now, I need to clean up and remove > debug/dev code from what I have, also I want to complete the mktest.pl > conversion before I do so. However at this point in time the reworked > stuff works just fine with a moderately patched mktest.pl. > > If anybody is willing to put up with pre-alpha grade stuff to review what I > have then just ask and you will recieve. > > I hope my efforts in this direction meet with your comunal approval and > that by doing this I havent upset anybody. > > Cheers, > Yves good luck, Abe -- Amsterdam Perl Mongers http://amsterdam.pm.org #! /usr/bin/perl tie *STDIN, JAPH => qw(Just another Perl hacker);print while <STDIN>; sub JAPH::TIEHANDLE {$~=shift;bless \($^=join$/,@_),$~} sub JAPH::READLINE {$/=$^=~s/^(.+)\n?//?$1:undef;$\=$^eq""?"\n":" ";$/}
