WIP: Avoid VT-100 output from build.sh

[rainersigwald]

This is an attempt to work around/isolate the cause of intermittent
hangs in the CLI official build. Since the hang appears to be happening
when emitting a control code to STDOUT, let's try to set STDOUT up in
such a way that things mostly don't try to emit control codes to it.

Related:

• Intermittent hang in .NET CLI build msbuild#3563
• https://github.com/dotnet/corefx/issues/31559

[rainersigwald]

@mmitche @livarcocc I'd like to try this as a possible workaround/information-gathering approach for the intermittent CLI hang. Ideally we wouldn't have to check it into a mainline branch and could just spin many builds of a private branch to test just this change. Is that possible?

Note that I based my change on the last time build.sh changed, which was long ago, so it should cleanly merge to any branch but isn't a good codebase to actually test using--we'd need to merge it with a more recent HEAD.

[livarcocc]

@rainersigwald given that we have been hitting this issue in CI a lot, I would just try a few tests here on this PR itself. Just please, mark WIP in the title so that we don't accidentally merge it.

[tmds]

dockerrun.sh is passing a -t argument to docker run. That will allocate a terminal and cause control codes to be emitted.

[rainersigwald]

dockerrun.sh is passing a -t argument to docker run. That will allocate a terminal and cause control codes to be emitted.

You're right, that's resistant to the stdout of the process that calls it. Boo. I'll bury it inside run-build.sh, I guess.

[mmitche]

@dotnet-bot test this please

We see this so much that it's been getting hard to get a build through.

[mmitche]

@livarcocc What do you think of these failures? Related?

[rainersigwald]

@mmitche I think they'll be fixed by #10147. Can we spin up a branch to run some official builds through, since this failure doesn't happen in CI?

[mmitche]

@rainersigwald Just push this branch to the internal remote and I'll take care of it.

[rainersigwald]

@mmitche I rebased onto #10147 and pushed to internal branch dev/raines/cat-build-output.

[mmitche]

@rainersigwald So far I haven't seen the failure with this in the build, though let's wait a little longer since not all the builds have made it through to the CLI.

[mmitche]

@rainersigwald I haven't seen any errors in the typical places (RHEL, Musl) though I have seen a few in arm64. Not sure whether it is the same issue. It may be worth merging this.

[nguerrera]

Do we plan to take this or can this be closed now?