terminal
terminal copied to clipboard
chore: Update to TAEF 10.93.240607003
I'm running this PR through the PGO build as well -- i will merge when it is done. @lhecker @zadjii-msft I need a re-review on the last push
This test failure on ARM64 is durable (it keeps failing) and nonsensical.
https://github.com/microsoft/terminal/blob/d3a18b904123b95a7b5a1ea2307f064cf3e28c94/src/buffer/out/ut_textbuffer/TextAttributeTests.cpp#L116-L127
Line 127 is failing, because GetColor is returning #0C0C0C instead of #00FF00. We just constructed the attribute with green as the background index above.
I need to look at this on an actual ARM64 device...
Also it only fails in release.
Figures.
I am starting to suspect an ARM64 code generation issue.
attr.GetBackground().GetColor(...) optimizes down into a direct load of index 0 from the color table.
This is after the entire attribute is const-initialized to for sure be IsRGB.
Here's a less red-yarn-and-thumbtacks image of the comparison and data flow
Did we ever file an internal issue on the ARM codegen thing here?
I believe it's this one: https://developercommunity.visualstudio.com/t/Bad-ARM64-codegen--maybe-during-constant/10583372
got it, so that's https://task.ms/dd/1969044 internally. For the sake of triaging the PR backlog, I'm gonna move this to a draft (since we can't merge till the compiler bug is fixed)
we can't merge till the compiler bug is fixed
The compiler bug reproduces in complete isolation, without TAEF. I don't think this PR caused it, it only found it.
If this arm64 build passes, I will reupdate and ingest taef 10.92
The compiler bug may only be fixed by VS 17.11 (!)
Banished back to draft status
It's still broken! AHHAHAHA
/azp run
Pull request contains merge conflicts.
/azp run