Closed op closed 11 years ago
Can you check if #35 fixed the memory leak please?
This PR allocates memory that is assigned to a static variable and is never freed. I have not researched what global_cleanup does but perhaps that is the appropriate place to free the allocated memory.
Without the freeing, the memory may be considered "leaked" by various tools.
And now we are talking about a performance optimization for the single malloc.
A simpler implementation that performs no heap memory allocation is #39. But it requires that LIBCURL_VERSION be a compiler #define, if it ever changes to a variable libcurl will break.
That looks very similar to the original change.
Nice to see that pycurl lives again! Here's a patch I submitted to Debian[1] too.
[1] http://bugs.debian.org/709226