Build: fix pre-compiled header files for MSVC/Ninja

[mistafunk]

Description of Change(s)

• Fix use of pre-compiled header when using MSVC with Ninja
• Fix potential compilation errors with Ninja caused by concurrent writes to PDB files when building with multiple cores.

• I have verified that all unit tests pass with the proposed changes

• I have submitted a signed Contributor License Agreement

[meshula]

can we also revise the comment to not say Windows, but rather "msvc with msbuild"

[mistafunk]

Maybe just remove this comment - or do the homework and measure this properly :-)

I did a quick stopwatch comarison... msvc/ninja with PCH took about 5min and without PCH about 6min... so we could also just say "... are a win with MSVC ..."

[pmolodo]

On another note - I've been using sccache with windows msvc USD builds recently, and been pretty happy with it so far. Haven't done any time comparisons, but rebuilds feel faster - particularly if you switch branches and come back, so a lot of stuff has updated timestamps but hasn't actually changed.

[meshula]

thanks for noticing this

[mistafunk]

thanks for noticing this

You are welcome. I took another look and found the actual bugs. There was a missing directory and wrong cmake command order when PCH was enabled. I reused this PR and force pushed.

[mistafunk]

A note for CLion users, do not get confused (like I did) by output like this in the CMake window:

Problems were encountered while collecting compiler information:
	C:\Users\<xxx>\AppData\Local\Temp\compiler-file7019753171835842146: fatal error C1083: Cannot open include file: 'pxr/usd/_ar/pypch.h': No such file or directory

This is an artifact of the CLion "intellisense" and not related to the USD build system...

[meshula]

Thanks, this looks ready for an internal review.

[mistafunk]

Btw, I tried to run the tests, there are over 100 failing out of 600 - is this exepected on dev?

[jesschimein]

Filed as internal issue #USD-9931

[jesschimein]

/AzurePipelines run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[sunyab]

@mistafunk Our internal CI is currently showing no failures, so I would not expect anything like that. I also can't imagine anything specific to this change that would be causing that either.

Do you have an example of a failing test you could share here? I'd bet they're all failing for the same reason.

[pmolodo]

I'll close my PR3141, since it looks like this fixes that and does a bit more as well!

[mistafunk]

I cannot reproduce the linux build failure locally... is the pipeline timeout too low?

[DDoS]

There is also #3120 that fixes this same issue.

[adskWangl]

There is also #3120 that fixes this same issue.

Thank you and we'll close #3120 as that is for the same issue.