On 02/25/2016 05:46 PM, Jakub Jelinek wrote:
Ah, I've been looking for something that would set die_perennial_p, but actually you just set die_mark later on instead for those. So IMHO the right fix is just handle all the ops that could directly or indirectly contain references to other DIEs, rather than just handling the 3 you have there.
Agreed! (as per the patch I sent)
Going to bootstrap/regtest this on x86_64-linux and i686-linux now. Is this ok for trunk if it passes testing?
I guess this was not for me as I’m not maintainer, but: this looks completely fine to me. :-) Thanks!
-- Pierre-Marie de Rodat