Re: midlife crisis what crisis?
I just read that same thread - I fundamentally disagree.
Linus complaint was:
* a patch in the audio code had been accepted by Mauro that changed the error return codes of a function, this changes the interface which other applications could rely on.
* this patch in the audio code broke a kernel release candidate, the fault was not picked up by testing tools Mauro also had responsibility for maintaining.
* having broken a RC & received explanation of the problem error code he questions whether the userspace code is wrong rather than his code -- he allowed the interface to be changed!!
* Linus suggests he should shut up before he embarrasses himself further.
* then Mauro keeps talking .. have a funny feeling this has happened before. There is no excuse. Revert to the previously returned return codes. And larger changes require discussion first.
* Mauro suggests that it is ok to make a change to an interface (which is in production) to gain consistency with something else. This is absolutely wrong, once the interface is set it cannot be changed, other applications are relying on it. This is the source of Linus frustration - Mauro hasn't learnt the first rule of kernel maintenance - you can't randomly change shit others could be relying on. If you do & something breaks it is the kernel's fault.