[Stdlib][ABI] Resolve ABI FIXME #25 #6246
Conversation
| // CHECK-NOT: .loc | ||
| // CHECK: .loc {{[0-9]}} 6 {{[0-9]}} | ||
| // CHECK: .loc {{[0-9]}} 6 {{[0-9]}} prologue_end |
There was a problem hiding this comment.
@adrian-prantl This change seems sketchy to me.
There was a problem hiding this comment.
Interestingly enough, this is something that I was recently looking at. Something has changed and when building swift against upstream-with-swift on clang/LLVM, it seems that the prolgue_end marker is no longer emitted. The MF does see the FrameSetup markers on the prologue instructions, but when the line record is constructued, we ignore the FrameSetup MIFlag and just drop the prologue tracking on the floor.
Also, it seems that for some reason this test is marked as XFAIL on Linux. Why is that?
There was a problem hiding this comment.
Judging from the version control history, this testcase was XFAILed some time during the Linux bringup and hasn't been looked at since.
@compnerd: could you file bugs for both of these issues and CC/assign it to me?
|
@swift-ci please test. |
|
Build failed |
|
This should address the build failure: #6247 |
|
Thanks @slavapestov. I'll take this opportunity to do what I promised in July and write a real doc comment for @jrose-apple |
|
Let us know when/if you want that radar closed :) |
Process arguments are now fetched per-platform
CodaFi
force-pushed
the
the-farmer-refuted
branch
from
f68848b
to
fe00dc7
Compare
|
@swift-ci please smoke test. |
|
IIRC we still needed |
|
|
|
All right then. Thanks, @CodaFi! |
|
⛵️ |
|
@slavapestov You can close the radar now! |
I introduced this in #3108 because we no longer require this SPI.
Resolves rdar://problem/19696522