added native windows threading support #101
Conversation
Could you clean up so it's easy to read the diff? Thanks! |
|
Ok, I dos2unixed the files so the diff should be more readable. I also fixed the few stupid mistakes causing the build issues on the other platform. |
| @@ -183,6 +80,167 @@ enum { | |||
| SEMKIND_INTERNAL | |||
| }; | |||
|
|
|||
| #if (defined(LIBETPAN_REENTRANT) && defined(HAVE_PTHREAD_H) && !defined(IGNORE_PTHREAD_H)) || !defined(LIBETPAN_REENTRANT) | |||
|
|
|||
| static int mailsem_internal_init(struct mailsem_internal * s, | |||
There was a problem hiding this comment.
Could you unindent the content of #ifdef?
|
Is that better? The spacing comments confused me a bit, I think that the issue was that I was using hard-tabs so I changed them to soft tabs. |
|
Going through the diff, there's still lots of indentation changes. |
|
Could you go through "Files Changed" section and make sure there's no odd diff. |
|
Could you give me a general overview of how you would like the indentation to be? E.g. use soft tabs not hard tabs, do not indent sub #defines, etc. |
Usually, those rules are not too strict. The things to fix are the changes in spacing and indentation. |
| r = pthread_mutex_init(&s->lock, NULL); | ||
| if (r != 0) | ||
| goto err; |
|
How is that? (For some reason the diffs dont show the contents of the file. Is there a delay on the diffs?) |
|
@chacham15 |
|
btw, why did you introduce IGNORE_PTHREAD_H? |
|
Because it is possible to have (and use) pthreads under mingw. Therefore, I added that option to allow one to explicitly build with only windows threads. |
|
IGNORE_PTHREAD_H is never defined. How do you set it? |
|
I merge most of it in e67180b. |
|
apologies, I added it because I knew that it would be useful, not because I personally would need it at this time. I went back and added the proper definition to configure.ac. |
|
I'm closing this pull request now since I merged manually. |
|
However, could you check if it builds properly for you on Windows? |
I wanted to compile libetpan under mingw and not use the mingw-pthread library so I implemented all the changes necessary. The only thing that I am a bit uncertain about (although in practice it seems to work fine) is a result of the lack of a static initializer in windows of a mutex. Therefore, I added some initialization stuff in mailstream_ssl.c at mailstream_ssl_init. The code itself will run initialization only in one thread. The only problem is that if two threads run it simultaneously, the thread that did not do the initialization might continue on to use the thread primitives while they are being initalized by the other thread. Naturally, this is only a problem for this compilation of windows, but so far I see no problem. If you have any ideas of how to fix that minor problem I would be grateful. In any case, the changes I have made here do seem to work for mingw with native windows threads if you would like to import back to the main branch.