A great step forward.. better than what I'd done (I was trying to make minimal changes rather than overhauling) and matches what phobos has, for the most part.

One concern.. the unit tests are based on release mode, which is rather problematic.  Release mode disables asserts -- ie nothing is really tested. :)

Sean, I'll re-look at my changes, but between this one and the change you're already started on (getting the sync primitives to build) are all I had, I think.

Later,
Brad

Brad Roberts wrote:
> A great step forward.. better than what I'd done (I was trying to make minimal changes rather than overhauling) and matches what phobos has, for the most part.
> 
> One concern.. the unit tests are based on release mode, which is rather problematic.  Release mode disables asserts -- ie nothing is really tested. :)
> 
> Sean, I'll re-look at my changes, but between this one and the change you're already started on (getting the sync primitives to build) are all I had, I think.
> 
> Later,
> Brad
> _______________________________________________
> D-runtime mailing list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime

Hrm.. the docs for the compiler suggest that -unittest turns asserts on, so there are contradictory arguments -release and -unittest in action at the same time.  I guess I'll need to actually test to see what _really_ happens, but I have been fighting with getting testing some other changes that I haven't submitted to anyone to look at yet.

Confused,
Brad

Which makefiles?

Andrei

Brad Roberts wrote:
> A great step forward.. better than what I'd done (I was trying to make minimal changes rather than overhauling) and matches what phobos has, for the most part.
> 
> One concern.. the unit tests are based on release mode, which is rather problematic.  Release mode disables asserts -- ie nothing is really tested. :)
> 
> Sean, I'll re-look at my changes, but between this one and the change you're already started on (getting the sync primitives to build) are all I had, I think.
> 
> Later,
> Brad
> 

The ones you emailed me a week or two ago.  I finally checked them in yesterday.

As for unit tests... that's kind of weird.  I think the correct thing is to not have either -release /or/ -debug set, since you don't want the debug blocks activated, nor do you want to lose asserts (or contracts, I suppose).  I struggled with this for druntime, and finally decided that the "debug" build would not set any flags, including debug, and the "release" build would set the release flag.  The idea being that the build should be used for someone to debug their own code, so debug blocks within the lib itself weren't appropriate, but debug symbols, contracts, etc, were.  However, I think I subsequently forgot about this decision and made the unittest build link the release lib.

Andrei Alexandrescu wrote:
> Which makefiles?
> 
> Andrei
> 
> Brad Roberts wrote:
>> A great step forward.. better than what I'd done (I was trying to make minimal changes rather than overhauling) and matches what phobos has, for the most part.
>>
>> One concern.. the unit tests are based on release mode, which is rather problematic.  Release mode disables asserts -- ie nothing is really tested. :)
>>
>> Sean, I'll re-look at my changes, but between this one and the change you're already started on (getting the sync primitives to build) are all I had, I think.
>>
>> Later,
>> Brad
>>
> _______________________________________________
> D-runtime mailing list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime