[release/6.0] Single-file apps should not use copy-on-write mapping of the .exe on windows.

[github-actions]

Backport of #59033 to release/6.0

/cc @agocke @VSadov

Fixes: #55405 (App published as single file fails to run in Windows 10.0.17763 containers)

Customer Impact

Regression from .NET Core 2.1 and 3.1 causing failure when a single-file app runs in a Windows 10.0.17763 container.

(This is a relatively old 2018 RS5 build, but it is an LTS so it will be around for a while)

Testing

Manually verified in an actual container. Regular Unit tests.

Risk

Very low.

We do not benefit from copy-on-write part of the mapping on Windows since we do not do in-situ json parsing due to UTF8/UTF16 differences. We basically request COW for consistency with Unix. Somehow it triggers unexpected "access denied" failure for superhost-style single-file apps on this particular build/container.

Since we do not do in-situ parsing on Windows, a read-only mapping is sufficient and we should request just what we need.

[dotnet-issue-labeler]

I couldn't figure out the best area label to add to this PR. If you have write-permissions please help me learn by adding exactly one area label.

[jeffschwMSFT]

Approved. Please take a look at the failing CI and we can take for consideration in RC2.

[ghost]

Tagging subscribers to this area: @agocke, @vitek-karas, @VSadov
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Backport of #59033 to release/6.0

/cc @agocke @VSadov

Fixes: #55405 (App published as single file fails to run in Windows 10.0.17763 containers)

Customer Impact

Regression from .NET Core 2.1 and 3.1 causing failure when a single-file app runs in a Windows 10.0.17763 container.

(This is a relatively old 2018 RS5 build, but it is an LTS so it will be around for a while)

Testing

Manually verified in an actual container. Regular Unit tests.

Risk

Very low.

We do not benefit from copy-on-write part of the mapping on Windows since we do not do in-situ json parsing due to UTF8/UTF16 differences. We basically request COW for consistency with Unix. Somehow it triggers unexpected "access denied" failure for superhost-style single-file apps on this particular build/container.

Since we do not do in-situ parsing on Windows, a read-only mapping is sufficient and we should request just what we need.

Author:	github-actions[bot]
Assignees:	-
Labels:	area-Single-File
Milestone:	-

[VSadov]

bunch of legs cancelled with

The agent did not connect within the alloted time of 45 minute(s).

is there a way to un-cancel?

[janvorli]

@VSadov rerunning the legs usually helps. If not, then closing / reopening the PR would re-run all the legs.

[VSadov]

I did not see an option to rerun on any leg, so i wondered if i would need to close/reopen.
Now it seems unstuck again.

[danmoseley]

@MattGal here is another case where there are a bunch of canceled ones after close/reopen a PR. This time they didn't clean themselves up. I'm guessing @VSadov didn't cancel them.

[MattGal]

@MattGal here is another case where there are a bunch of canceled ones after close/reopen a PR. This time they didn't clean themselves up. I'm guessing @VSadov didn't cancel them.

Based off what was happening (machines failing to "phone home" after provisioning) this is what I would expect.

I was just on a call with @ulisesh and we narrowed this down to a setting about interactive login for the agent. While some repos that do UI-testing will be adversely affected by this, un-setting the value seemed to immediately allow machines to provision at scale (even when their provisioning scripts "fail"; this seems to be a red herring found on the way to the answer).

The interactive login setting is off now, so clicking through to AzDO and doing "Rerun failed jobs" should now work (I tried it here to check)

[danmoseley]

@mdh1418 @steveisok is this a known issue?

               adb: failed to install /datadisks/disk1/work/B2C50999/w/ABD90955/e/System.Text.Json.SourceGeneration.Roslyn3.11.Tests.apk: Failure [INSTALL_PARSE_FAILED_BAD_PACKAGE_NAME: Invalid manifest package: bad character '1']
                 

https://helixre8s23ayyeko0k025g8.blob.core.windows.net/dotnet-runtime-refs-pull-59194-merge-cc1b7cc32c704243a1/System.Text.Json.SourceGeneration.Roslyn3.11.Tests.Attempt.3/1/console.a68ff716.log?sv=2019-07-07&se=2021-10-06T19%3A17%3A32Z&sr=c&sp=rl&sig=469W5Gc9%2BNsfQxKCfVL%2B%2Bn97GUC0SF5e3nDGPcLcT7o%3D

[steveisok]

Yes, spotted it earlier today.

We use the name of the project as the name of the android app. The numbers in the name violate android naming rules. We'll be able to address them tomorrow.

[danmoseley]

Any idea how it got through the system though? This project was added by #59074 and as far as I can tell, all its runtime-staging lanes passed.

[steveisok]

@danmoseley You won't see the test failures unless you look on the azdo side.

https://dev.azure.com/dnceng/public/_build/results?buildId=1361380&view=results

[VSadov]

All tests green, finally.

[agocke]

Ironically, I think the right branch for this is now -rc2

[danmoseley]

logged both failures: should be good to merge.

[danmoseley]

@steveisok

@danmoseley You won't see the test failures unless you look on the azdo side.

Should I expect a yellow on the Github side though? I mean, what is the indication that someone should go look at the azdo side?

[steveisok]

I wish that was the case, but from what I understand is that we can't due to AzDo limitations. With the staging pipeline, we have to go in and look.