Enable transcription of native commands on non-Windows#4871
Enable transcription of native commands on non-Windows#4871daxian-dbw merged 7 commits intoPowerShell:masterfrom
Conversation
There was a problem hiding this comment.
Also need to check that PS error stream is empty.
There was a problem hiding this comment.
Actually, there's no reason to follow that pattern, I'm rewriting it as a scriptblock and any errors should just show up in the test run.
There was a problem hiding this comment.
This can be made a static check so that we do it only once. And _scrapeHostOutput can be renamed to s_supportScreenScrape (or something similar).
There was a problem hiding this comment.
Why do we want redirectInput to be true by default? Even if we are transcribing and screen scrape is not supported, we shouldn't need to always redirect the input.
There was a problem hiding this comment.
And form line 442 below, it indicates that we transcribe for native command only if _runStandAlone is true. Now it looks like you are assuming we can do transcribe regardless of _runStandAlone. That seems not right.
There was a problem hiding this comment.
Yes, redirectInput shouldn't be true by default. Will fix.
As for _runStandAlone, my understanding of the code is that if not running standalone, output is redirected and will be captured in the transcript. However, if running standalone, the output is screen scraped and put into the transcript. So it seems that regardless if running standalone, if screen scraping is not supported, the only way to transcribe the output is to redirect it.
There was a problem hiding this comment.
Suggestion: this.Command.Context.EngineHostInterface.UI.IsTranscribing is used at 3 places in this class. Maybe we can have an instance field to cache the value? Like _isTranscribing = this.Command.Context.EngineHostInterface.UI.IsTranscribing in the constructor?
There was a problem hiding this comment.
from line 444 below, it looks like a HostException also indicates The host doesn't support scraping via its RawUI interface ...
There was a problem hiding this comment.
I think the catch for HostException was a mistake. Going through the code in the debugger, that exception isn't thrown.
There was a problem hiding this comment.
By putting the call to GetBufferContents in the constructor, it's possible that an arbitrary exception will be thrown from GetBufferContents in the constructor. When that happens, it would be wrapped to an ApplicationFailedException properly prior to this change, but after this change, the NativeCommandProcessor would just fail to be constructed, and thus I guess the error would look obscure too. So, I'm not sure if it's appropriate to put this in the constructor ...
@SteveL-MSFT Could you please explain a little why redirecting the output/error would support transcribing in case that screen scraping is not supported? (just curious how it's done in that case) |
|
@daxian-dbw from what I can tell in the code, anything that gets redirected gets transcribed anyways |
722b161 to
c72288d
Compare
There was a problem hiding this comment.
I think setting redirections for transcribing purpose should be done in the method CalculateIORedirection, and IMO it should be done at the end of that method -- after the existing logic of that method has run. For example, in case of running on the server side, redirectInput should always set to be true. Code like that should run even though screen scraping is not supported.
There was a problem hiding this comment.
Makes sense, will fix.
There was a problem hiding this comment.
I think there is a change of behavior here. Assuming _isTranscribing == $true and screen scraping is supported,
- before the change,
_scrapeHostOutputis true only if_runStandAloneis true, so this block will run only the native command is running standalone. - after the change, even if it's not running standalone (i.e. standard input/error redirected), this block of code still will run when transcribing, and in such case,
_startPositionis not even initialized.
There was a problem hiding this comment.
Maybe remove the else? The else if (Platform.IsWindowsDesktop && IsConsoleApplication(this.Path)) block doesn't necessarily set redirectOutput and redirectError to true.
There was a problem hiding this comment.
And please add a comment that explains why we need to set output/error redirection in such case.
There was a problem hiding this comment.
These changes now can be reverted.
|
@anmenaga Can you please take a look again? |
015ca77 to
f769df1
Compare
f769df1 to
b63b374
Compare
|
@markekraus the test failures are indicating the WebListener didn't start within the timeout. Is this something you're familiar with? |
|
@SteveL-MSFT News to me. I will take a look. |
|
@SteveL-MSFT This appears to be related to the change in this PR. It looks native execution inside a job is broken. WebListener runs expected behavior (working in beta.7): Actual behavior: |
|
@markekraus thanks! I'll take a look. The problem is that when run as a job obviously there's no console, so some of the console calls even on Windows throws other exceptions (and it's not just the Host.HostException from the old code) when the code is checking if screen scraping is supported. Added a test to test start-job with native command to cover this. @daxian-dbw @anmenaga can you take another look? |
…ot supported since screen buffer reading isn't supported, default to using redirected output
b63b374 to
2ff64d8
Compare
| catch (Exception) | ||
| { | ||
| // screen scraping not supported on non-Windows or as job | ||
| } |
There was a problem hiding this comment.
I think this check can still be done only once -- we do this check in CalculateIORedirection and only if _runStandAlone is true. Here the the code I propose:
private void CalculateIORedirection(out bool redirectOutput, out bool redirectError, out bool redirectInput)
{
...
if (NativeCommandProcessor.IsServerSide)
{
redirectInput = true;
redirectOutput = true;
redirectError = true;
}
else if (Platform.IsWindowsDesktop && IsConsoleApplication(this.Path))
{
// On Windows desktops, if the command to run is a console application,
// then allocate a console if there isn't one attached already...
ConsoleVisibility.AllocateHiddenConsole();
if (ConsoleVisibility.AlwaysCaptureApplicationIO)
{
redirectOutput = true;
redirectError = true;
}
}
_runStandAlone = !redirectInput && !redirectOutput && !redirectError;
// !!!!! NOTE !!!!! -- Unchanged until here
if (_runStandAlone)
{
if (s_supportScreenScrape == null)
{
// Call GetBufferContents() to check and set s_supportScreenScrape to true or false accordingly.
}
// if screen scraping isn't supported, we enable redirection so that the output is still transcribed
// as redirected output is always transcribed
if (_isTranscribing && (false == s_supportScreenScrape))
{
redirectOutput = true;
redirectError = true;
_runStandAlone = false;
}
}
}So the check of screen scraping support is done only if we are running the native command as standalone application.
In case of running it in a Job, I think the standard input and error are already redirected, so nothing needs to be done.
There was a problem hiding this comment.
I see. I'll try that.
Transcription was relying on reading the screen buffer to record output from native commands.
This resulted in an unhandled exception calling an unimplemented api on non-Windows.
Fix is to default to using redirected output if reading the screen buffer is not supported.
Fix #1920