Lawrence Greenfield wrote:
>
> Date: Sun, 10 Feb 2002 17:20:03 -0800
> From: Erik Steffl <[EMAIL PROTECTED]>
> [...]
> > We currently use yacc (or bison) to parse Sieve scripts; the bytecode
> > work isn't currently expected to change this. We basically report the
> > parse errors that Yacc gives us.
>
> it often does report it but not always. and test program ignores them!
>
> The test program ignores what?
the test (in sieve subdirectory) ignores sieve error messages, just
prints error numbers which is not very helpful. the error messages are
visible when debugging the test program but they are not printed. one
such example is when one message matches two reject rules, the error
message says something about two matches for reject but test only prints
number (one can look up the number in one of the include files but it
doesn't say anything).
that's a separate problem from the parse errors, those are missing
completely (some), you can see comments in sieve (yacc file, IIRC) that
the error message should be set but it's not...
> [...]
> 2. if sieve does not have good error reporting then how do you debug
> the programs that generate sieve files?
>
> Sieve's formal grammar is not particularly difficult; it really isn't
> very hard to write a program that correctly generates scripts.
it can be quite tricky when you don't see what's going on. e.g. the
above mention two reject rules, or my another mistake when I used mail
message as sieve file (it says parse error on line one, doesn't say
which file it was processing, where exactly the error was etc.) - silly
errors but can take quite some time to figure out by just staring at the
monitor. If the error was reported properly (givin enough information)
both these problems would be non-issues (solved in a matter of seconds).
>
> getting the information from cyrus (what's going on, errors etc.) is
> the worst part of cyrus, that's not only my opinion, all other people
> that I discussed the configuration of cyrus had the same problems.
>
> I agree error reporting could be improved. It's all a matter of time.
>
> note that this is NOT a documentation problem, docs could be better as
> well but they are generally fairly OK, this is the problem of programs
> reporting their status.
>
> I'd be willing to add some error messages at least in sieve (that's
> were it's needed most, IMO), how's the policy on accepting patches?
>
> You send 'em to [EMAIL PROTECTED], we usually apply them or
> at least try to communicate what we don't like.
>
> Major features can take some time or might be shunted to the contrib
> directory (like drac-auth).
I would just like to have error/status messages added, I guess I'll
try that when I have time and send the patches, see how that works...
thanks,
erik
