windowsbatch-filecmdpushd

implicit "popd" on batch exit

Is there a way to undo all pushd at the end of script. What I have is:

pushd somwhere
rem doing stuff
goto end

:end
popd
goto :EOF

What I'd like to have is:

setlocal
pushd somwhere
rem doing stuff
goto :EOF

But that doesn't work, the directory stays "pushd". Another try where I can't say how many pushd will occur would be:

set CurrentDir=%CD%
rem doing a lot of pushd and popd
pushd somwhere

:POPBACK
if /i not "%CD%" == "%CurrentDir%" popd & goto POPBACK

But that looks like I can easily get stuck at :POPBACK. Adding a counter to exit that loop is a bit uncertain at which dir I end up.

In that context I'd like to know if a can see the stack of pushd directories. The only thing I found was $+ in prompt $P$+$G which adds a '+' for each pushd directory e.g. D:\CMD++>. But I can't see a way how to make use of that.

EDIT: I just noticed there is something wrong with the $+ of prompt that confused me. The setlocal in the example above actually makes the script return to the initial directory, but the prompt shows a '+' indicating a missing popd.

D:\CMD>test.cmd
D:\CMD>prompt $P$+$g
D:\CMD>setlocal
D:\CMD>pushd e:\DATA
e:\DATA+>goto :EOF
D:\CMD+>popd
D:\CMD>

EDIT: Due to the hint to the difference between local environment and directory stack I expanded the example form above, pushing to C:\DATA before calling the script. the script returns to where it has been called from (C:\DATA). So far so good. The first popd shows no effect as it removes the last pushd from stack but not changing the directory, since the directory change has been undone by the (implicit) endlocal. The second popd returns to initial directory as expected.

D:\CMD>pushd C:\DATA
C:\DATA+>d:\cmd\test.cmd
C:\DATA+>prompt $P$+$g
C:\DATA+>setlocal
C:\DATA+>pushd e:\DATA
e:\DATA++>goto :EOF
C:\DATA++>popd
C:\DATA+>popd
D:\CMD>
Solution

  • Two points here:

    1. Setlocal/Endlocal commands save/restore current directory!

    This means that when Endlocal command is executed, the current directory is returned to the point it has when the previous Setlocal command was executed. This happens no matter if the Endlocal is explicitly executed, or it is implicitly executed by the termination of the called subroutine (via exit /B or goto :EOF commands).

    This behaviour could be used to solve your problem, because all changes of the current directory are reverted when endlocal command is executed. That is:

    setlocal
cd somwhere
rem doing stuff
goto :EOF

    However note that this behavior is not related to pushd/popd commands! You can not mix the use of this feature with pushd/popd commands. You would need to complete several tests until the mechanism of the interaction between these two features be comprehended.

    This is an undocumented behavior discussed here.

    2. POPD command sets the Exit Code when there is no previous PUSHD

    As described below Exit Code management section at this answer, there are some commands that does not set the ErrorLevel when an error happens, but sets an internal value we could call "Exit Code". This value can be tested via the && || construct this way:

    command && Then command when Exit Code == 0 || Else command when Exit code != 0

    POPD is one of this type of commands (described under Table 5), so we could use this feature to execute POPD several times until there is not a matching PUSHD previously executed:

    @echo off
setlocal

echo Enter in:
for /L %%i in (1,1,%1) do echo %%i & pushd ..

echo Leave out:
set "n=0"
:nextLevel
popd || goto exitLevels
   set /A n+=1
   echo %n%
   goto nextLevel
:exitLevels

echo Done

    Bottom line: a simple popd && goto POPBACK line solves your problem...