Save and Restore WorkdirDepth when removeFile

[robbycandra]

When remove recovery file. (in the root dir), workDirDepth will be set to 0.
Therefore we need to save and restore workDirDepth when removing recovery file.

This will fix Power Loss Recovery bug when Resume from second power loss.

[thinkyhead]

This works if it happens to leave the contents of the work dir hierarchy (from SD navigation) alone. But if the recovery file is ever stored anywhere other than the root directory this will break.

Some breaking up of cardreader.cpp has been done already to simplify basic file open. If we keep tuning it to the needs of Marlin we ought to end up doing some other tweaks, like having the SD current dir tree stored separately from the general diveToFile dir tree, and so on.

That said, another approach to this PR would be to add a flag to diveToFile to tell it not to mess with the work dir tree, and this will prevent it from overwriting the work dir position and tree data in all cases.

Let's give that a try since diveToFile is definitely used in places other than navigation via the LCD.

[robbycandra]

This is simple solution that works for now.

Ideally, we should have 3 work dir:

• One for LCD navigation.
• One for GCode file selection.
• One for Power Loss Recovery file and maybe Another System File (in the future).

But the Melzi user will not like it. We stucked we stucked by 128K memory board.

[thinkyhead]

This is simple solution that works for now.

The time it takes to type comments should be used to type code. So I will do that now.

[thinkyhead]

Ideally, we should have 3 work dir:

The only real "work dir" needed for core functionality is the dir of the running file so that it can execute sub-programs. Above that, a working directory for the LCD file navigation. These both need the history array because in the first case you need to be able to go back to the level of the caller. And for LCD, the data is needed to go "back" up the tree while navigating. During printing you can't navigate the media from the LCD so once a job is going the history is re-applied. So in these cases, a single navigation tree does suffice. The DOS path as a string would also be a fine proxy for the array of SDFile objects we now use. We can quickly dive to sub-folders from the root every time, if that were needed; it wouldn't add much extra time at all to dir reading.

[robbycandra]

@thinkyhead , i think current solution is still not correct.
The purpose to Save and Restore the workdirDepth is to prevent change in the workDirDepth when delete /PLR file (in the root dir -> workdirDepth = 0).

Without save and restore if we remove /PLR then

• If we update_CWD than we current workdirDepth will be lost
• If we dont update_CWD than /PLR cannot be deleted because marlin search in wrong directory.

This without save and restore, problem occur when :

• We Print a file in sub-dir
• The Power outage happened twice.
  because in the first outage, the record of workdir in PLR file still correct. After Resume, marlin start recording worng workdir in the PLR. At second outage, resume will not start because marlin search the printed gcode file in wrong directory . Marlin search in root dir when actually the file located in the sub dir).

[robbycandra]

For workDir system in the LCD and Gcode (/HOST), i still think that we need 2 workDirDepth system, because they can be run in seperate branch. Therefore, not only the depth is different, but the tree-branch can be different too...

[thinkyhead]

The concept of "current working directory" (workDir) has not existed except as part of the LCD navigation of the SD card. However, this workDir is respected for card.removeFile or card.openFileRead if a relative path is given (no / at the start). For normal usage via G-code you must always give the full root-relative path to the thing you want to delete.

/PLR cannot be deleted because marlin search in wrong directory.

In the case of deleting "/PLR", as you can see the slash at the start of the pathname means the path is root-relative. The workDir is not relevant to anything that gives a root-relative path. Again, the diveToFile effect on workDir has not been altered by the last change, except to now update it for any openFileRead (called by M23, M32, and printingHasFinished for sub-programs).

Prior to this change diveToFile always changed the path array and never changed workDir. Now, as before, the LCD interface changes workDir and the path array whenever it needs to cd into a folder or cdup out of a folder. Re-inserting the SD card will set workDir back to root, but this is not problematic, since only SD-based print jobs can run sub-programs anyway.

[robbycandra]

I think i need to re-read the code. Thanks for your explanantion.

[thinkyhead]

After Resume, marlin start recording worng workdir in the PLR.

• The full path to the file being printed is stored in the PLR file.
• One of the changes I made to diveToFile ensures that workDir is pointed to the folder containing the file when it resumes printing.

[robbycandra]

Hi @thinkyhead , I confirm that this code is OK and working. You are really good. Thanks.