Pete Robbins wrote:
On 16/12/06, Jean-Sebastien Delfino <[EMAIL PROTECTED]> wrote:
Under revision r487731 I added some code to Logger.cpp to check the
length of log messages to avoid buffer overruns, in particular when
logging big XML documents or SDOs. I also changed the buffer allocated
on the stack from 4096 to 256 chars as most log messages are smaller
than that, and bigger buffers are now allocated from the heap instead.
I had to write this differently on Linux and Windows (vsnprintf on Linux
and a combination of vnsprintf and _vscprintf on Windows). Could
somebody try to compile and test this Mac-OS X as well? To test it, run
any scenario that flows big XML docs (PythonWeatherForecast is a good
one) with TUSCANY_SCACPP_LOGGING=2, and make sure that the log does not
contain garbage and that the entire XML docs returned by the Web Service
are logged .
I'll test it out on Mac but I'm having trouble building the REST
extension.
The tuscany_sca_rest_reference and tuscany_sca_rest_service both link
with
tuscany_sca_rest_interface. This causes duplicate definitions of the
symbol
tuscany_sca_extension_initialize... which sort of makes sense. We may
need
to split the tuscany_sca_interface library up into the part that is the
extension and ... the rest if the other libs need to link against this. I
take it this is not a problem on Linux?
Does it mean that if you dynamically load DLLs A and B for example,
which both link to C, then C is getting loaded twice? Is this specific
to what we do with our extensions or a more general problem on the Mac?
I think that there's also a related problem on Windows, when we
dynamically load an extension DLL A linked to another extension DLL C,
we need to have the directory containing DLL C on the Windows PATH
(ironically we don't need A), and it's really confusing and not convenient.
I think we need to look at our three Linux, Mac-OS, WIndows platforms
and try to find a consistent scheme that works well on all of them.
--
Jean-Sebastien
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
