Andreas,
    see Dean Roddey's message as of 22 May 2001 20:06:47 UTC for an
explanation of how to stop the memory leaks.

(The following applies to UNIX, not the Win32 and Mac transcoders -
picked up by "projects", not runConfigure/Makefiles.)

Re. the transcoders - you can choose which transcoding service you
want to use by using the "-t" option in runConfigure. Currently you
can choose either of "icu", "Iconv390" (IBM), "Iconv400" (IBM) and
"native" (which is the Iconv service used by default).

Every implementation of the platform utilities has pre-processor
statements (#ifdef, #else and so forth) to instantiate the proper
transcoding service at runtime, depending on the option chosen when
generating the Makefiles.
(For your box, have a look in XMLTransService::makeTransService() in
$XERCESCROOT/src/util/Platforms/Linux/LinuxPlatformUtils.cpp)

You will see that if you do not specify -t at all (or use the
default -ticonv) you will instantiate a new IconvTransService object
as the result of calling XMLTransService::makeTransService() (by the
way, this particular constructor happens to be empty and this call
just allocates a new object and grabs the reference).

You were looking to far down in IconvTransService.cpp, get back a few
lines and you will see that you were in section "The protected virtual
transcoding service API". Have another look at IconvTransService::* if
you are interested in the details.

Anyway, if you're not interested it does not really matter - Dean has
the right answer to your core problem! :)
    /Martin

--
Martin Kalen
Software Engineer
TODAY Systems, Inc.
http://www.todaysystems.com.au/
Tel +61-3-9536 3900 - Fax +61-3-9536 3901

----- Original Message -----
From: <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Wednesday, May 23, 2001 2:23 AM
Subject: Antwort: RE: DOMString and international characters


>
> Hi all,
>
> I'm also using Xerces on a Linux platform and thus concerned by the
leaking
> ISO-8859-1 characters greater than 128. I think the bug is already
in the
> bug tracking system.
>
> Now I've found out, that among the Transcoders ( in
src/util/Transcoders )
> there are realizations for Win32, IBM390, IBM400, Mac and a standard
Iconv,
> which is probably used by Linux OS. IconvTransService has a function
> XMLTranscoder* makeNewXMLTranscode( const XMLCh* const encodingName,
... )
> which is not fully implemented ( it just returns a 0 ). This would
explain
> the behaviour. I thought to build a new TransService, perhaps called
> LinuxTransService, derived from IBM390, which seems to be fairly
close to
> what we expect.
>
> My question is now, if there is already someone implementing the
Linux -
> Transservice? Or is there perhaps a realizations around?
>
> Bye, Andreas.



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to