curl-library
RE: Character code conversions
Date: Mon, 16 Jul 2007 11:43:20 +0200
Thanks for your answer
> If I read between the lines, it seems you yourself consider the
alternative 2 mentioned above to be the proper choice...
Yes it is: the only drawback is the adding of several non-standard
wrappers procedures to support character conversion of arguments.
I'm currently working on it. It will be ready in a few days (coding done
100%, testing 10%).
> I just want this work to not spill over and "damage" platforms that
don't need or want the conversions, ...
The only "featured" modifications I had to apply to the "common code" is
about curl_easy_setopt() (in fact, Curl_setopt()): string arguments are
considered static, this does not allow easy handling of dynamically
allocated converted strings; I therefore implemented an "allocation
flags map" in the UserDefined structure to support freeing of strings
when the structure is released. These changes do NOT modify the
behaviour of regular curl_easy_setopt() calls. But they may allow you to
offer additional options or entry points in the future, to implement
dynamic release features...
Writing a wrapper to curl_formadd() was a pain in the ass...
I'll send you a big patch in a few days: the small updates to the
regular code and the new directory packages/OS400 and its contents. This
will allow use of libcurl on OS400, both in ASCII and single-byte EBCDIC
variants, and from C, C++ and RPG/ILE (the most used language on OS400).
Currently, it works for http, https, ftp (with or without SSL), ftps and
dict. Supporting ldap would require to suppress the dlopen() stuff: I
know you have some plan to do it, but we'll talk about ldap later.
Patrick
Received on 2007-07-16