[cfe-commits] [PATCH] Add -fobjc-direct-class-refs

[John McCall]

On Jun 5, 2012, at 10:24 AM, Jonathan Schleifer wrote:
> Am 05.06.2012 um 19:16 schrieb David Blaikie:
>> Except that once it is added you can rely on it, assuming you never
>> intend to use an old compiler (& why would you?)... and then Clang is
>> stuck supporting it for all those users who've come to depend on it.
> 
> I for one check if the flag is available or not. Code using my code usually gets the compiler flags from my config script. So removing it has zero impact. One should never assume a flag is always there anyway - and it has already happened before that Clang removed options.
> 
> Anyway, I forked Clang now, as I have never seen an opensource project refusing a simple option so fiercely while allowing thousand others, and then brazenly demanding to rewrite the whole approach how ObjC runtimes are handled just for a single option. Due to the fork, that option will be in use soon anyway. And forking, renaming clang to something like oclang and getting it in distributions seems 1000 times easier and less work than actually getting a patch into Clang.
> 
> I have tried hard, I invested hours into getting into Clang code, I even added a new runtime class, a new runtime flag etc. as was requested, but you guys keep demanding more and more - I give up. I'm not going to spend days for a single option to rework your architecture for you. I consider this blackmailing to say "Either you rework our architecture for us, or you won't get your option you need so desperately". I tried hard, but you really don't let someone any option besides forking.
> 
> The fork is at https://github.com/Midar/clang for those who are interested. Currently I only applied my original patch, I will rename it to oclang later and submit it to distributions.

Sigh.  I'll try to hack together a -fobjc-runtime flag next week during WWDC, if someone doesn't beat me to it.

John.