Add emission of Frontend parseable-output in WMO jobs. #58422
Conversation
artemcm
force-pushed
the
ParseableWMOOutput
branch
from
3ead41b to
41afdc5
Compare
|
@swift-ci please test |
artemcm
force-pushed
the
ParseableWMOOutput
branch
from
41afdc5 to
bef7970
Compare
|
@swift-ci please test |
| const char *const Delim = ""; | ||
| std::ostringstream JoinedDiags; | ||
| std::copy(PrimaryDiags.begin(), PrimaryDiags.end(), | ||
| std::copy(AllDiagnostics.begin(), AllDiagnostics.end(), | ||
| std::ostream_iterator<std::string>(JoinedDiags, Delim)); | ||
|
|
||
| emitFinishedMessage(llvm::errs(), | ||
| mapFrontendInvocationToAction(Invocation), | ||
| JoinedDiags.str(), r, Pid - idx, ProcInfo); | ||
| return false; | ||
| }); | ||
| JoinedDiags.str(), r, OSPid, ProcInfo); | ||
| } |
There was a problem hiding this comment.
Nit: I notice that this code is essentially duplicated between the two branches, just with slightly different parameters to emitFinishedMessage(). Do you think there's a way you could combine them?
lib/FrontendTool/FrontendTool.cpp
Outdated
| // Outputs for Primary Inputs | ||
| for (const auto &input : PrimaryInputs) { | ||
| auto OutputFile = input.outputFilename(); | ||
| Outputs.push_back(OutputPair(file_types::lookupTypeForExtension( | ||
| llvm::sys::path::extension(OutputFile)), | ||
| OutputFile)); | ||
| } |
There was a problem hiding this comment.
Is this correct for WMO invocations that produce a single main output? In that case, I believe the first input has the main output + supplementary outputs associated with it, and the remaining inputs have empty outputs. Should we be using forEachOutputFilename and forEachInputProducingSupplementaryOutput here?
There was a problem hiding this comment.
I believe this will be equivalent to what the code does now already. Right below this code-block we iterate over all the inputs again and gather supplementary outputs per-input. So while the iteration may be redundant, we should not miss any outputs. I've updated the test with an additional supplementary output just to make sure it is consistently shown in both WMO and non-WMO.
There was a problem hiding this comment.
Right yeah, we won't miss any outputs, but I believe we'll be writing out empty outputs for all but the first input in a WMO invocation with multiple inputs and a single main output. Is that expected?
There was a problem hiding this comment.
Ah, of course, now I see it. Will fix, thank you.
artemcm
force-pushed
the
ParseableWMOOutput
branch
from
bef7970 to
4348fdd
Compare
|
@swift-ci please test |
artemcm
force-pushed
the
ParseableWMOOutput
branch
from
4348fdd to
9e787f2
Compare
|
@swift-ci please test |
…shed messages on a per-primary basis. Which means that `-frontend-parseable-output` had no effect in contexts where no primary inputs are specified, such as WMO. This change fixes that by also emitting began/finished messages when no primary outputs are specified. Resolves rdar://91999048
artemcm
force-pushed
the
ParseableWMOOutput
branch
from
9e787f2 to
9d2aecb
Compare
|
@swift-ci please smoke test |
|
@swift-ci please smoke test Windows platform |
|
@swift-ci please smoke test Linux platform |
Frontend emitting parsable output, as implemented, emitted began/finished messages on a per-primary basis. Which means that
-frontend-parseable-outputhad no effect in contexts where no primary inputs are specified, such as WMO. This change fixes that by also emitting began/finished messages when no primary outputs are specified.Resolves rdar://91999048