Re: sourceforge pserver is known bad. Should VCP work around?
Barrie Slaymaker <[email protected]> Wed, 18 Feb 2004 20:00:23 -0500
| Newsgroups | gmane.comp.version-control.revml |
|---|---|
| Message-ID | <[email protected]> |
On Wed, Feb 18, 2004 at 01:47:02PM -0800, Matt Attaway wrote: > > Question 1: Is it acceptable to break a VCP conversion just because > > the underlying command emits stderr? > > > Sadly, probably not. A lot of apps, Perforce included, occasionally > spew non-critical error messages to STDERR. It would be nice if there > was some way to tell VCP to ignore certain messages so that over time > we can build up a list of non-critical messages. That's how it's done now, grep the sources for "stderr_filter" and you'll see some regular expressions that are used to match stderr and ignore or react to matching output. It's a short list and I'm worried about fragility in light of the CVS weirdness we just tri[pped over at sourceforge. > Likewise, sometimes really important error messages come out on > STDOUT. We don't have to figure out what all to ignore right now, as > long as there is a mechanism to do so going forward. Nod. > > Question 2: Should there be a user prompt before we bomb out of the > > conversion? > > > That could be worthwhile. That may be another way to deal with non-critical > errors, especially for this first release. It will give users a handy way > to ignore pointless error messages while we sort out which ones we > should ignore. > > Question 3: Should there be a command line and .vcp file option to allow > > the user to set policy? > > > Probably not. A flag to tell it to plough through no matter what might be > useful. However, one of the interesting bits about people who are > converting is that they usually know one system very well, and the > other hardly at all. It would be hard for them to make any intelligent > decision. The more smarts we can stick into VCP about what to ignore > and what to listen to, I think the better off we will be. > > Reading my response there, I realize I'm not sure what you mean by > a policy. What do you envision? Options to control VCP's response, perhaps like: --ignore-all-unexpected-errors --ignore-errors-matching=.*foo.*bar - Barrie