-
Thanks Greg. It looks fine.
2008-10-27 02:55:20 UTC in UPthreads
-
Hi Greg,
Sorry for my late reply. I cannot access the CVS of SourceForge in my environment, so there might be some inconsistency between CVS and my tree.
Which mail address should I use? patacongo at users.sourceforge.net ? Will it accept attachment?
Chuang.
2008-10-20 05:19:24 UTC in UPthreads
-
Hi Greg,
I have implement a queue for all queue operations in upthreads. It is based on realloc() and works somewhat like vector in C++.
There are two drawbacks in this implementation:
First, the interface is not clean still. Users access items in the queue through an array. This may be a problem if we want to change the design later.
Second, the queue cannot be initialized...
2008-10-13 09:01:55 UTC in UPthreads
-
Hi Greg,
I mean currently upthreads-0.1.4/*/*.[ch] have not yet been processed by indent.sh. If I modify some files, use indent.sh and then diff, I will still get a patch with unrelated difference. If this is not acceptable, then maybe it will be better to release a version that all source files have already been processed by indent.sh.
If you don't mind, then we can use indent.sh each...
2008-09-25 09:13:45 UTC in UPthreads
-
combo.c and mqueue.c fail on mq_send() while CONFIG_MEMORY_ALLOCATOR is undefined because of not enough CONFIG_MQUEUE_MSGSIZE.
Fix two warnings while CONFIG_MQUEUE_PREALLOCATED and CONFIG_SEM_PREALLOCATED are defined.
2008-09-23 08:59:10 UTC in UPthreads
-
Sure, I agree. The key point is consistency but not details. I mentioned that because I read the comments in the script.
upthreads-0.1.4 does not yet applied the indent rule. Will you release a new version for it?.
2008-09-10 15:01:50 UTC in UPthreads
-
Hi Greg,
I think this script will be easier to use. We can modify many files at the same time:
$ ./tools/indent.sh libpthread/*c libpthread/*h libmqueue/*c libmqueue/*h include/*h
btw, with regard to the trailing */, how about -cdb?.
2008-09-10 04:01:15 UTC in UPthreads
-
Hi Greg,
In this patch, there is no more pthread_return(). A thread yields cpu by calling pthrad_schedule() and the scheduling is done in the same context. The scheduler context and scheduler stack are no more necessary.
Previously, to perform a context switch, 2 longjmp() calls are required - longjmp() to scheduler and longjmp() to new thread. Now, all the scheduling works are done in...
2008-09-09 08:55:26 UTC in UPthreads
-
in signaler_thread(), nerrors is incremented while succeeded.
2008-09-09 07:55:18 UTC in UPthreads
-
Thanks for your appreciation.
This issue is not urgent to me. It does not happen in my application - just for curious. Besides, I had done a larger change on scheduler (remove the need of dedicated scheduler stack) and need some time to clean up to make a patch. It is not bad to defer it.
2008-05-13 02:19:26 UTC in UPthreads
