Michael Peters <mpet...@plusthree.com> writes: > Steffen Schwigon wrote: > >> That could work. The TAP streams could start with the same >> "t/xyz.t ..." line as prove already does. Or some other more >> unambiguously detectable delimiter line. > > The minute we start adding delimiters between TAP streams we again > risk introducing non-TAP into the output. Right now TAP ignores > anything it doesn't recognize as non-TAP but that amorphous blob of > what-is-not-TAP can change and thus break any saved single file TAP > archives. > >> That still raises the need that the remaining TAP output is guaranteed >> to be non-TAP, which brings us back to the start of the discussion. > > Not even considering the delimiter we already have non-TAP since > multiple streams will mean multiple plans, restarting test numbers, > etc which will still make it non-TAP.
*sigh* I know. When I currently collect TAP for my archive and recognize it as beeing a piped prove output I already cut it into sections, currently on prove's "t/xyz.t ..." lines. I know, it's not really solving the topic generally enough for prove... Hm, looks like neither the "-" file for printing to STDOUT nor even the single file will make sense under those conditions. Isn't there a silver bullet for such problems, like introducing heavy-weight XML protocols everywhere? :-) Kind regards, Steffen -- Steffen Schwigon <s...@renormalist.net> Dresden Perl Mongers <http://dresden-pm.org> German Perl-Workshop <http://www.perl-workshop.de>