[Bug 57137] New: endlocal changes CWD if pushd/popd are called before

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57137

Bug ID: 57137
           Summary: endlocal changes CWD if pushd/popd are called before
           Product: Wine
           Version: 9.16
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: cmd
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Run test.bat, got output

Z:\home
Z:\home\huangqinjin
Z:\home

test.bat:

@echo off

pushd "%~dp0.."
cd
popd

setlocal 
cd
endlocal
cd

[Bug 57137] startlocal/endlocal changes CWD if pushd/popd are called before

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57137

huangqinjin  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|endlocal changes CWD if     |startlocal/endlocal changes
                   |pushd/popd are called       |CWD if pushd/popd are
                   |before                      |called before

[Bug 57137] startlocal/endlocal changes CWD if pushd/popd are called before

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57137

Ken Sharp  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |source, testcase

[Bug 57137] startlocal/endlocal changes CWD if pushd/popd are called before

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57137

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download

[Bug 57137] startlocal/endlocal changes CWD if pushd/popd are called before

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57137

Eric Pouech  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #1 from Eric Pouech  ---
(not sure I'm going to work on this in the short term; so writing some notes if
someone is willing to look into it)

that will require more testing to investigate the impact on internal
structures...

some quick look:
- current directory for all drives is saved by setlocal on windows (not only
current one)
- and all drives are restored by endlocal (on windows)

I don't think tweaking the setlocal/endlocal structure to store all drives as
dedicated fields is the correct approach
IMO, we should rather use the =<Drive>: env variables; they are partly
supported in builtin cmd, but will likely need more work...

Indication: 'echo %=c:%' on native shows the default dir for drive C (doesn't
work yet on builtin) (see [1] for more details)

likely to be assessed (non exhaustive list):
- how are the env variables inherited/created on cmd.exe creation (wild guess:
inherited when they exist, and set to =<drive>:=\ when not present at startup)
- how they are passed to child process (wild guess: simply passed as part of
current env)
- impact on CD/CHDIR builtin commands (wild guess: likely update env var
accordingly, partly done, requires some more tests)
- likely low impact on the rest of the code as RtlGetFullPathName_U already
handles them

-Eric

[1] https://devblogs.microsoft.com/oldnewthing/20100506-00/?p=14133