michaelselehov wrote:

> Reverted in #206678.
> 
> Maybe we should do #205822 after all?

@nikic 
Looking at the revert: am I right that the compressed bundle header fields 
(FileSize etc.) are serialized/read in host-native byte order rather than fixed 
little-endian, so they're only correct on little-endian hosts? It looks like 
this was already latent before my PR - your #205822 (and the s390x 
fatbin-coff-compress.test crash) predates my merge - and my change just made it 
worse by relying on FileSize for bundle splitting and adding more 
endian-sensitive tests.

I'd like to re-land my fix on a correct foundation. Since #205822 wasn't 
merged, would you object if I combine that change with the re-landing PR?

https://github.com/llvm/llvm-project/pull/205587
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to