[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: 2 build failures on win32




Thanks, this is great news!!

Did you have to do anything special?  Other than

1. build OpenSSL
2. define HAVE_TLS in portable.nt
3. Add OpenSSL libraries to the link dependencies of the executables that need it.


Anything else?

--Kervin

Jens B. Jorgensen wrote:
Kervin L. Pierre wrote:

1. back-perl build is broken due to the include of XSUB.h. patch attached.

2. Build fails in sasl.c, line 180. The build is configured without TLS, therefore the Connection structure does not have the 'c_is_tls' member. maybe this line has to be wrapped with HAVE_TLS, as the definition is in slap.h?

PS. Has anyone tried to build on win32 with TLS enabled? Is it possible?


Yup. I have a working build that we're using right now with TLS enabled (actually we only use TLS connections: '-h ldaps:///'). I did this by first building openssl in windows which works. Note that I did all this with VC++, *not* cygwin/gcc/mingw32.

As a side note, a while ago I wrote an SSL socket class which uses windows SSPI rather then openssl. Figuring out how to do this was a pain but it works reliably. It would be nice to get code into the source that would use SSPI rather than openssl. I would be happy to contribute such an effort myself but I don't have any spare time to do this.

I wouldn't mind comparing notes with others doing win32 builds. I have a working build as I mentioned with a custom backed that hits an oracle database. It works but is very slow for some reason. When I'm running it in a console window with debug output I can see that it is pausing in weird places, eg. after the incoming socket connection is established. Watching the behavior from the window I don't think the slowness is related to any external (database, ssl) performance problem but is rather due to something about how the socket handling is implemented. BTW I have a multi-threaded build.