racke at linuxia
Nov 14, 2009, 2:07 AM
Post #2 of 2
(623 views)
Permalink
|
Stefan Hornburg (Racke) wrote:
> Hello, Interchange developers!
>
> We found on the IRC meeting on Tuesday:
>
> --snip--
> The character set conversion for file uploads happens in
> Vend::Server::parse_multipart before we know which catalog the file upload
> belongs to. David and Racke agreed that we store the character set for
> each uploaded file in $CGI::file_encoding first and do the conversion in
> Vend::Dispatch::open_cat just before the call to open_database().
> Functions which write the data in $CGI::file to the file system need
> to pass the corresponding value in $CGI::file_encoding to
> Vend::File::writefile.
> --snap--
>
> Further investigation yielded that every CGI post parameter is subject
> to this character set conversion _and_ that we don't know the character
> set for the catalog.
>
> I looked into the source code and found no reason to parse CGI
> input ahead of calling Vend::Dispatch::open_cat. The only thing which
> is used from CGI is $CGI::script_name.
>
> My proposal is to postpone parsing of CGI parameters and run it
> from inside open_cat.
>
Please see attached patch. It needs more testing and maybe some
polishing, but in general it should work fine :-). I didn't try
if it makes problems with mod_perl mode.
Regards
Racke
--
LinuXia Systems => http://www.linuxia.de/
Expert Interchange Consulting and System Administration
ICDEVGROUP => http://www.icdevgroup.org/
Interchange Development Team
|
enc.patch
