Egregoros · Phoenix Framework

Post

Remote status

Adrian Chadd <verified.png>

In another case of "you broke something, then you just disabled it" i give you:

OpenSSL "Fixing" PPC64 HW AES crypto handling (save/restore VSX registers) which broke PPC64 bigendian so they just said "oh this is broken, disable it" rather than asking the original developer to fix it.

So, the VSX fix:

https://github.com/openssl/openssl/commit/8f24a148e6f112045cf110c3e3f4e2415d9c11fe

.. which they obviously didn't test on PPC64.

Then someone noticed it broke PPC BE:

https://github.com/openssl/openssl/issues/29845

which was "Fixed" by:

https://github.com/openssl/openssl/commit/1d19060b55b3ab206cf1399220a4bdb5bd7c2d42

Replies

Replying to @erikarn@mstdn.social

Adrian Chadd <verified.png>

@erikarn@mstdn.social remote

So I'm going to reach out to the original bug filer (q66, who's here on fediverse) to see if we can figure out how to PROPERLY fix it so we don't just disable hardware crypto handling on big endian platforms.

This btw is how a platform dies the death of a thousand cuts - laziness.

(also note the bug 100% fired in QEMU, so "I don't have hardware" isn't a good enough reason. You can run the test suite in a QEMU VM!)

feld reposted

Adrian Chadd <verified.png>

@erikarn@mstdn.social remote

In another case of "you broke something, then you just disabled it" i give you:

So, the VSX fix:

https://github.com/openssl/openssl/commit/8f24a148e6f112045cf110c3e3f4e2415d9c11fe

.. which they obviously didn't test on PPC64.

Then someone noticed it broke PPC BE:

https://github.com/openssl/openssl/issues/29845

which was "Fixed" by:

https://github.com/openssl/openssl/commit/1d19060b55b3ab206cf1399220a4bdb5bd7c2d42