Build Master: Scheduling

[deadalnix]

On Thursday, 14 November 2013 at 20:30:40 UTC, Dicebot wrote:
> On Thursday, 14 November 2013 at 20:27:23 UTC, Rob T wrote:
>> Bug fixes should be as frequent as possible. To understand 
>> why, just try and find a good reason to artificially hold back 
>> a bug fix.
>
> Code breakage. DMD has good amount of accepts-invalid bugs.

Right now I have 15 tests failing in SDC test suite because of 
DMD messing up with closures.

This is certainly not an option to wait 6 month that way. Accept 
invalid is certainly not something you can stick in patches 
release, but invalid codegen is.

I fairly bummed by code breakage being raised as a excuse to not 
do anything. As a matter of fact, I have to work with my own 
homebrew version of DMD half of the time. One release out of 2 
has introduced unacceptable regression.

Now I'm ready to hear that people don't want to spend time 
publishing a minor version of DMD because they don't have time or 
whatever. But this breaking code bullshit must stop.

Every single release of DMD break my code, 1 release out of 2 
break it by introducing regression, not because of accept-invalid 
(which I'm happy to fix).

We need these minor release. Anyone that have a reasonably sized 
D codebase knows it. In fact I'm pretty sure that anyone having a 
reasonably sized D codebase is using some homebrew version of DMD 
most of the time, as waiting 6 month is usually not an option.