Re: PR/32004, tree-ssa caused in/out asm constraints to often need reloads

[Eric Botcazou <ebotcazou at adacore dot com>]

> His patch did not even touch local-alloc.c at all or anything which
> could have caused this directly (rtl.def).  So this is a latent bug.

It might well be, but did you investigate?

> The main reason why he applied it without any approval is based on the
> rule that the bug is a regression and it was approved for the trunk.

You don't put a brand new pass on a branch that is in maintenance mode, 
especially if you cannot disable it, precisely because you don't know
how it will interact with the rest of the compiler.

If Paolo had requested approval for the 4.1 branch, I'd have opposed.

-- 
Eric Botcazou