Re: [xml] [long] bugreport: HTTP resource escaping/unescaping fails with quoted URLs

[Daniel Veillard <veillard redhat com>]

On Fri, Dec 13, 2002 at 09:25:30AM +0100, Stefano Zacchiroli wrote:
> On Thu, Dec 12, 2002 at 04:43:49PM -0500, Daniel Veillard wrote:
> >   Not a big deal just indicate the current version in the report
> <snip>
> >   I hadn't any time for it. Bugzilla is there to prevent me from
> > forgetting about issues which occurs for mail stuff
>
> Ok, I will submit a bug report also to bugzilla so that we have a
> 'record' of it.

  okay, thanks that helps,

> >   It wasn't looking trivial, not fun to do or test, and it still in
> > the folder. Provide a patch if you want to garantee fast processing.
>
> Ok. Anyway I wasn't asking for fast processing, just for a bit of
> feedback that you have read the mail and you have considered it to be a
> bug.

  Hum, I'm getting lazy. I didn't had the time to go through the
code path and have no reason to doubt it might be wrong, so I hadn't
anything specific to say about it.

> > > I hope that my analysis is correct ...:
> >
> >   Maybe, I don't know by heart the full set of code, and part of this
> > has been really nasty e.g. handling Windoze file names.
>
> Bleeargghh ...
>
> >   then spend time to learn more about it
> <snip>
> >   and provide a patch. Make it checked for Windows paths and verify it
> > passes asll libxml2 and libxslt regression tests. Then you issue
> > will get fixed promptly if you don't suggest to make silly things
> > like changing APIs
>
> Ok, I can try to have a look at it, but I definitely have no time to
> study the libxml/libxslt internal structure; so it can help if you can
> give me some additional information.
> Mainly I would like to know:
> - if you remember the architectural reason why you chose to unescape the
>   URI in xmlParserInputBufferCreateFilename

  because precisely that was shared by all the underlying I/O layers.
Retrospectively that was probably a mistake as you pointed out.

> - what do you mean by regerssion testing (is enough running all the
>   test* executables available in libxml directory or someone else is
>   needed? do those tests includes tests on windoze filenames?)

   make tests in both libxml and libxslt trees once rebuilt and
linked with the new code.
I can't test windows filename, I don't have Windows.

> - assuming that I will be able to make it work, do you like my proposal
>   of keep on using the received, still escaped, URI until it reachs
>   nanohttp.c?

   Sounds fine but you will need to unescape on the others I/O handler
to preserve the existing behaviour on those. All this is contained in
xmlIO.c and xmlParserInputBufferCreateFilename(), that a (very) small
subset of the code.

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/