Hi,
Someone in the Subversion team made our C tests run in parallel. With that
our buildbot found a race condition in the Windows specific
APR_DECLARE_LATE_DLL_FUNC() implementation.
The current code is
[[
#define APR_DECLARE_LATE_DLL_FUNC(lib, rettype, calltype, fn, ord, args,
names) \
Hi,
On Windows apr_arch_misc.h checks for the 'WINNT' define, which was
unconditionally defined on Windows before the CMake build. But now it is no
longer defined. This enables some additional dynamic load operations that
could fail in some cases (See Race condition in
On Thu, Dec 5, 2013 at 9:13 AM, Bert Huijben b...@qqmail.nl wrote:
Hi,
On Windows apr_arch_misc.h checks for the ‘WINNT’ define, which was
unconditionally defined on Windows before the CMake build. But now it is no
longer defined. This enables some additional dynamic load
G.
On 12/5/2013 9:23 AM, Gregg Smith wrote:
On 12/5/2013 6:48 AM, Jeff Trawick wrote:
On Thu, Dec 5, 2013 at 9:13 AM, Bert Huijben b...@qqmail.nl
mailto:b...@qqmail.nl wrote:
Hi,
On Windows apr_arch_misc.h checks for the ‘WINNT’ define, which
was unconditionally defined on Windows before
On Thu, 05 Dec 2013 09:24:24 -0800
Gregg Smith g...@gknw.net wrote:
Once all the 9x stuff is gone, we can drop the define.
Precisely. And that would be a version-major/minor change, imho.
In the interim, simply -D (ugh... /D it) WINNT. Problem solved. Leave
it there for someone to discover
On Thu, 5 Dec 2013 15:01:05 +0100
Bert Huijben b...@qqmail.nl wrote:
I think the dll load function should be converted to a more stable
pattern, that properly handles multiple threads. And perhaps we
should just assume a few more NT functions to be alsways there
instead of loading them