Re: checklink: checklink losing on member-only documents (and/or unhelpful diagnostic message)

On Sunday 19 April 2009, C. M. Sperberg-McQueen wrote:

> I am confused.  The diagnostic message is quite right:  you DO need
> W3CACL access to check the links in the document.  And I have it.
> If the link checker doesn't have a password and userid, it is welcome
> to ask me for one (as for example the validator currently does, with
> success); it isn't doing that.
>
> I think the link checker used to be able to check W3C member-only
> documents without problem (I have hard-coded links for checking WG
> documents I need to check regularly); if it is still intended to be
> able to check member-only documents, it appears to me that something
> is broken.

Right, asking for username/password was broken because some HTTP headers were 
output too early.  Fixed now in CVS and the development instance at http://qa-
dev.w3.org/wlc/checklink

http://dev.w3.org/cvsweb/perl/modules/W3C/LinkChecker/bin/checklink.diff?r1=4.155&r2=4.156&f=h

> (And for that
> matter, the expression given is not a regular expression as that term
> is defined by any book on formal languages I have ever read --
> I assume the line noise at the beginning is some Perl-inspired
> excrescence.  Perhaps you should say "only to hostnames
> matching the regular expression (in Perl notation) ..." ?)

The problem is that we really have no dependable info about how perl outputs 
the regex.  All we have at that point is a compiled perl regex, and it is not 
documented how such beasts are printed.  With current perl versions, it 
appears to be perl's "extended pattern syntax" but I don't think we can count 
on that being the case in all perl versions.

Anyway, the current CVS version no longer says anything about the trusted 
regexp if the hostname matched it, and makes the error message a bit clearer 
when it doesn't.

http://dev.w3.org/cvsweb/perl/modules/W3C/LinkChecker/bin/checklink.diff?r1=4.157&r2=4.158&f=h

Received on Sunday, 19 April 2009 11:16:59 UTC