Make StartInfo_TextFile_ShellExecute test more reliable

[adamsitnik]

edit: introduce a new, unique extension and use the registry to associate it with Notepad to be always sure that it's going to be used

Fixes #25823

[ghost]

Tagging subscribers to this area: @eiriktsarpalis
See info in area-owners.md if you want to be subscribed.

Issue Details

---

use notepad.exe in an explicit way (like the two other notepad tests do), don't rely on it being associated with .txt files

otherwise, things like notepad++ get opened (common on dev machines) and the test fails

Fixes #25823

Author:	adamsitnik
Assignees:	-
Labels:	area-System.Diagnostics.Process
Milestone:	6.0.0

[adamsitnik]

/azp run runtime-libraries-coreclr outerloop-windows

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[adamsitnik]

@Anipik @ViktorHofer how can I trigger the "Spanish" CI?

[jkotas]

otherwise, things like notepad++ get opened (common on dev machines) and the test fails

We should move the test to outer loop. It is ok for outer loop tests to have environment dependencies like having notepad.exe registered with .txt files.

[adamsitnik]

We should move the test to outer loop.

It's already an outer loop test:

runtime/src/libraries/System.Diagnostics.Process/tests/ProcessStartInfoTests.cs

Line 962 in 38c9b98

[OuterLoop("Launches notepad")]

It is ok for outer loop tests to have environment dependencies like having notepad.exe registered with .txt files.

I think that we should introduce a new, unique extension and use the registry to associate it with Notepad to be always sure that it's going to be used. PTAL at my most recent commit

[adamsitnik]

/azp run runtime-libraries-coreclr outerloop-windows

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[ViktorHofer]

@Anipik @ViktorHofer how can I trigger the "Spanish" CI?

Replied offline but to summarize, the "Send to helix" log shows which queues work was sent to:

Sending Job to Windows.81.Amd64.Open... 

 Sent Helix Job; see work items at https://helix.dot.net/api/jobs/547b1787-d975-4927-8d42-be6f276c4663/workitems?api-version=2019-06-17 

 Sending Job to Windows.10.Amd64.Server19H1.ES.Open... 

 Sent Helix Job; see work items at https://helix.dot.net/api/jobs/c034fbf8-a38e-48d6-acda-47ff9fe8f4dd/workitems?api-version=2019-06-17 

 Sending Job to (Windows.Nano.1809.Amd64.Open)windows.10.amd64.serverrs5.open@mcr.microsoft.com/dotnet-buildtools/prereqs:nanoserver-1809-helix-amd64-08e8e40-20200107182504... 

 Sent Helix Job; see work items at https://helix.dot.net/api/jobs/2bc9eedd-d9e5-4b3c-8680-f70681e0ceea/workitems?api-version=2019-06-17 

 Waiting for completion of job 2bc9eedd-d9e5-4b3c-8680-f70681e0ceea on (Windows.Nano.1809.Amd64.Open)windows.10.amd64.serverrs5.open@mcr.microsoft.com/dotnet-buildtools/prereqs:nanoserver-1809-helix-amd64-08e8e40-20200107182504 

 Waiting for completion of job c034fbf8-a38e-48d6-acda-47ff9fe8f4dd on Windows.10.Amd64.Server19H1.ES.Open 

 Waiting for completion of job 547b1787-d975-4927-8d42-be6f276c4663 on Windows.81.Amd64.Open

In this case, Windows.10.Amd64.Server19H1.ES.Open is the Spanish queue which already runs per PR.

[adamsitnik]

From the CI logs it looks like only System.Net.Http.Functional.Tests.TelemetryTest_Http11.EventSource_ConnectionPoolAtMaxConnections_LogsRequestLeftQueue has failed (not related) and the rest have passed

[danmoseley]

Thanks, this seems like a good experiment to try. Was there any evidence though that the txt mapping was missing/bad? IIRC we log the reg key and it always seemed correct, the launch still failed, which left me confused.

I've tried to stabilize this poor test for a long time.

[adamsitnik]

Was there any evidence though that the txt mapping was missing/bad?

No, but it was the only reason why it was not passing on my dev machine so I thought that it would be good to improve it.

@danmosemsft is it OK if I re-enable now and when it fails get back to it with new logs?

[adamsitnik]

FWIW it's hard to act based on the old logs like #22007 (comment) because they contain source code line number, and the file has been modified many times and the line number does not tell me much

[danmoseley]

Happy with whatever approach seems best to you.

[adamsitnik]

Happy with whatever approach seems best to you.

@danmosemsft Can I get approval from you so I can merge the PR?