Andy Armstrong wrote: > On 14 Mar 2007, at 07:29, chromatic wrote: >> The problem is that there's no way to tell that that information >> sent to >> Test::Builder->diag() is diagnostic information for the tests >> because once it >> goes out on STDERR, it could be anything. > > So we seem to have two reasonably sensible options on the table. I > don't think they're mutually incompatible. > > Ovid's 'only merge STDOUT and STDERR when in verbose mode' seems to > be workable with current Test::Builder. > > We should also push forward with machine readable diagnostics as a > formal part of TAP and have those show up on STDOUT along with all > the other TAP. > > Did I miss anything?
Nope, looks sane. This whole debate has revealed that formal TAP diagnostics needs more consideration then what is in the current proposal. The ultimate goal is all information to go on STDOUT, even the warnings and errors if possible, in a way that the TAP parser can tell them apart. There's currently no proposal to do all that. But we can go ahead with TH 3 now using Ovid's plan without worrying about that.
