Alexey Popov wrote: > Hi. > > As far as I understand, compilation of regular expressions in ACL's is > performed during config parsing. Therefore regexp can be compiled either > as extended or basic depending on position of the keyword > 'extendedregex' in the config file. > > Not sure what to do, but I think it should be documented or one should > put keyword 'extendedregex' in the beginning of sample config file. I agree. That's a substantial oversight in the greylist.conf manpage. > I dont think this switch is useful at all. I disagree.. one thing to consider is that many versions of milter-greylist were released without support for extended regex. So you have an existing userbase that have configurations that rely on the basic regex syntax. While this switch might be useless to you, it's very useful to folks who have existing configurations. And this is, unfortunately, a very recent change for milter-greylist. The feature was added (by me) on March 18, 2005. The fact that the parsing changes depends on where you put it is more of a side-effect of how the parser is implemented. That said, it does have the useful benefit of allowing people to "tack on" extended regex acls to the end of their existing config without rewriting. But you're right, this effect of the feature should be documented. >re_format(7) in FreeBSD calls > basic regex "obsolete" and says that it "mostly exist for backward > compatibility in some old programs". It's not just FreeBSD that calls it obsolete, even POSIX calls it obsolete. IMHO, even extended regex is obsolete, surpassed by the far more popular PCRE (perl compatible regex). However, a conversion to PCRE would be a bit more involved. In the interim, at least we have extended regex support for those who need it. Which is substantially better than the state of things back in March, when milter-greylist did not support anything but basic regex.
Message
Re: [milter-greylist] logical bug with extendedregex
2005-10-28 by Matt Kettler
Attachments
- No local attachments were found for this message.