On 21 June 2016 at 14:57, Sebastian Benoit <be...@openbsd.org> wrote: > Henning Brauer(hb-openbsdt...@ml.bsws.de) on 2016.06.21 13:11:16 +0200: >> * Stefan Sperling <s...@stsp.name> [2016-06-21 11:15]: >> > Generally, I would appreciate more detailed error messages from the pf.conf >> > parser. I recall several occasions where pfctl threw "syntax error" and >> > more >> > specific error reporting would have saved me some time with finding the >> > silly mistake I made. And in this case the ruleset loads successfully even >> > though, while parsing, we already know it's not going to work as >> > intended... >> >> true, that's shared by all yacc-style parsers, if the grammar doesn't >> match you just get syntax error without much of a hint what's wrong. > > the quality of the error message depends both on the language (some make > it harder to error out early) and on the parser. > > fail early is better. > >> however, the ruleset in this case does NOT load. >> >> <brahe@quigon> $ echo '"a macro with spaces"="foo"\npass from $a\ macro\ >> with\ spaces"' | pfctl -nvf - >> a macro with spaces = "foo" >> stdin:2: macro 'a' not defined >> stdin:2: syntax error >> > > sure, thats what the op reported > >> >> > Only as long as it doesn't make the parser code overly complex, of course. >> > But currently the balance is tilted too much towards terse error messages >> > for my taste. So I liked benno's first diff. >> >> it's just a tiny check indeed, which swings the "cost" (not in >> financial terms) vs benefit pendulum towards the benefit side, yes. > > thx. > > here is a diff for all daemons. > > ok for this? > >
Looks good to me. OK mikeb FWIW