File: android-deltas-sync/common.py

"""
=======================================================================
Code used by scripts run on both PC and phone.

See _README.html for license, attribution, version, and docs.

Dev notes:

1) This file is loaded by import, and expects callers to immediately
run its startup(), passing in a config_{pc, phone} module which has 
global settings copied and used here.  This is due to its Bash heritage;
Py scripts here are more-or-less direct translations for portability.

2) In config_*.py files, set PauseSteps=n to avoid having to press Enter 
here, and set CheckFilenames=n to avoid name-fixer interactions here.

3) You can also change ZIP here to point to a newer installed ziptools,
but don't generally need to; the version in Mergeall-source/ suffices.
=======================================================================
"""

import os, time, sys, shutil, subprocess, errno
from os.path import join

VERSION = 1.2
APPNAME = 'Android Deltas Sync'    # renamed Oct22 [1.2]

# Linux also means WSL on Windows, and Android;
# Cygwin is Windows, but using its own Python

RunningOnWindows = sys.platform.startswith('win')       # may be run by Cygwin
RunningOnMacOS   = sys.platform.startswith('darwin')
RunningOnLinux   = sys.platform.startswith('linux')     # includes WSL, Android
RunningOnCygwin  = sys.platform.startswith('cygwin')    # Cygwin's own python
RunningOnAndroid = any(key for key in os.environ if key.startswith('ANDROID_'))

# common indent width
indent = ' ' * 4

# catch control-c in input() and exit nicely
builtin_input = input
def input(prompt=''):
    try:
        return builtin_input(prompt + ' ')
    except KeyboardInterrupt:
        print('\nRun aborted by control-c at prompt.')
        shutdown(bad=True)

import builtins
builtins.input = input    # default to version here everywhere



#----------------------------------------------------------------------
# Setup: globals, logs dir, etc.
#----------------------------------------------------------------------


def startup(config, self):
    """
    -------------------------------------------------------------------
    Common startup code, called with either the PC or phone config 
    module, and the module containing this function.

    This copies the config module's names into this module's namespace,
    and adds two more names here--and all of these names become globals 
    both here and in the client scripts.  Hackish, maybe, but globals 
    are denoted by all uppercase (mostly), and this is better than the 
    original Bash translation, which ran this file's all-top-level code 
    in-line in clients' namespaces to emulate a Bash 'source <file>':
    
        from config_pc import *
        mydir = os.path.dirname(__file__)
        exec(open(join(mydir, 'common.py')).read())

    This worked until the post hook script did the same thing and broke
    outputs.  Some globals may be avoided by passing args to utilities,
    but this module's namespace avoids extra args (~6 in some cases).
    -------------------------------------------------------------------
    """
    global ZIP, DATE    # set here!
    print('%s %.1f' % (APPNAME, VERSION))

    # copy pc-or-phone config settings to my namespace
    for name in dir(config):
        if not name.startswith('__'):
            setattr(self, name, getattr(config, name))

    # check that main common vars are set before tweaking
    for config in ('LOGS', 'MALL', 'FROM', 'TO', 'STUFF'):
        try:
            getattr(self, config)
        except:
            print('Config "%s" is not set' % config)
            print('Exiting; please check the config file and rerun.')
            shutdown(bad=True)

    # expand user home (~, ~user) and env vars ($var, ${var}) in paths
    expander = lambda path: os.path.expandvars(os.path.expanduser(path))
    for config in ('LOGS', 'MALL', 'FROM', 'TO'):
        setattr(self, config, expander(getattr(self, config)))

    # path checks in all scripts; STUFF is also checked in some scripts
    for config in ('MALL', 'FROM', 'TO'):
        setting = getattr(self, config)
        if not os.path.isdir(setting):
            print('The %s path does not exist: "%s"' % (config, setting))
            print('Exiting; please check the config file and rerun.')
            shutdown(bad=True)

    # ensure logfiles folder in all scripts
    if not os.path.isdir(LOGS):
        if os.path.exists(LOGS):
            os.remove(LOGS)
        os.mkdir(LOGS)

    # fix Windows non-ASCII prints when 'script > file' (cygwin too?)
    if RunningOnWindows:
        os.environ['PYTHONIOENCODING'] = 'utf8'    # subprocs inherit

    # ziptools source-code folder (included in Mergeall)
    ZIP = join(MALL, 'test', 'ziptools')

    # logfile date stamp
    now = time.localtime()
    DATE = time.strftime('%y-%m-%d', now)   # e.g., 21-09-30 for Sep 30, 2021 (sortable)



#----------------------------------------------------------------------
# Utilities 
#----------------------------------------------------------------------



def announce(steplabel): 
    """
    -------------------------------------------------------------------
    A conditional start-of-step display: pause for enter if enabled.
    -------------------------------------------------------------------
    """
    print('\nSTART:', steplabel)
    if PauseSteps:
        try:
            builtin_input('Press enter/return to continue') 
        except KeyboardInterrupt:
            print('\nRun aborted by control-c before step.')
            shutdown()



def opener(message):
    """
    -------------------------------------------------------------------
    Display opening message, save run start time for use in closer().
    Package name+version are shown in startup(), before error tests.
    -------------------------------------------------------------------
    """
    global RunStartTime
    RunStartTime = time.perf_counter()    # relative seconds
    print('\n' + message)



def closer(message): 
    """
    -------------------------------------------------------------------
    Normal exit: show total elapsed time (all steps + user time), and
    pause for an Enter press to allow output to be read if a PC-side 
    script was run on Windows by an icon click (see shutdown() docs).
    -------------------------------------------------------------------
    """
    global RunStartTime
    elapsed = time.perf_counter() - RunStartTime    # float seconds

    h, m, s = (elapsed // 3600), ((elapsed % 3600) // 60), (elapsed % 60)
    print('\nFinished - total elapsed time (h:m:s) = %d:%d:%.0f' % (h, m, s))

    print(message)
    shutdown()



def shutdown(bad=False):
    """
    -------------------------------------------------------------------
    All exits: pause for an Enter press iff a PC-side script was run 
    on Windows by an icon click, else the console window is closed 
    before its output can be read.  Then, end the script run now with
    shell status 1 if bad (errors), or 0 if not bad (good/normal).  

    Caveat: as coded, this may require an Enter in some other Windows 
    contexts besides icon clicks, unless streams are redirected.  Its 
    PROMPT test suffices to keep click windows up, and does not require
    an Enter in Command Prompt, Cygwin, or some IDEs (e.g., PyEdit).
    But it does require an Enter in both PowerShell and the IDLE IDE.  
 
    The only way to do better seems sniffing parent processes, and 
    this is too complex and brittle.  There really should be a better
    way to detect this very common Windows use case by now...

    Update: Termux:Widget launches similarly erase the console window
    on script exit (and open a different Termux session).  To support
    this, test a ForceKeyToClose config-file setting on phone.  This is 
    preset to False; Termux:Widget is nice, but likely rare in the wild.
    This config also works on PCs, but is not needed for Windows clicks.
    See also _etc/termux-widget-shims/ for wrappers that make this moot.
    -------------------------------------------------------------------
    """
    if (ForceKeyToClose or                     # Termux:Widget?

       (RunningOnWindows    and                # Windows +

        sys.stdout.isatty() and                # console output +
        sys.stdin.isatty()  and                # console input +

        'PROMPT' not in os.environ)):          # clicked only? (maybe)

        builtin_input('Press enter/return to exit')

    sys.exit(1 if bad else 0)



def offerFixnamesRun(loglabel, target):
    """
    -------------------------------------------------------------------
    On PC, ask to run the nonportable-filenames fixer, in both initial
    copies and syncs.  This is nearly required on Linux and macOS for 
    interoperability with Windows, and FAT32, exFAT, and BDR drives.

    But don't ask on Windows: Windows disallows nonportables globally,
    and Cygwin's munged nonportables register as the munges instead of
    the nonportables if it's using Windows Python.  Cygwin's own Python 
    does translates munges to nonportables, but also sets sys.platform
    to 'cygwin' so it's not RunningOnWindows.  WSL munges too, but its
    sys.platform is 'linux'.  See _etc/examples/windows*/x* for demos.

    For convenience, this scrapes the fixer's last two report lines as
    a summary, and pulls out the number-found at the end of the last.
    This is both brittle and error prone, and requires forging user 
    input for the script's verify, but reduces output and interaction.

    Update [1.2]: this is now also run on the _phone_ by part 1 of the
    new export scripts, because it's not impossible that names have
    been created in Android app-private storage which will not work 
    on Windows and some drives.  This required adding CheckFilenames
    to the phone configs file too (it will loaded from there by the
    on-phone export script), as well as the target path parameter here
    (which is join(FROM, STUFF) on PC, and join(TO, STUFF) on phone). 
    Auto-detect of app storage is tough; set CheckFilenames as desired.
    -------------------------------------------------------------------
    """
    if CheckFilenames and not RunningOnWindows:
        print('\n'
	'It is recommended to run fix-nonportable-filenames.py before\n' 
	'propagating content from Unix to Windows, some Android\'s shared\n' 
	'storage, and proxy drives using the FAT32 or exFAT filesystems.')

        fixer = join(MALL, 'fix-nonportable-filenames.py')
        logto = join(LOGS, '%s--%s-fixnames-log.txt' % (DATE, loglabel))

        userreply = input('Run the name-fixer script in report-only mode (y or n)?')
        if userreply == 'y':
            print('Running name fixer')
            lasttwo = runpy(fixer, target, '-', 
                            input=b'y\n',
                            logpath=logto, 
                            tailing=(2, 'Name-fixer summary'))

            # end interaction now if no lines to change
            numfound = int(lasttwo[-1].split(':')[1].strip())
            if numfound == 0:
                print('No names to change')
                return

            userreply = input('Display name-fixer\'s full output (y or n)?')
            if userreply == 'y':
                print(open(logto, 'r', encoding='utf8').read())   

        userreply = input('Run the name-fixer script to fix filenames (y or n)?')
        if userreply != 'y':
            print('Name-fixer script not run')
        else:
            print('Running name fixer')
            runpy(fixer, target, 
                  input=b'y\n',
                  logpath=logto,
                  tailing=(2, 'Name-fixer summary'))



def verifyPCtoProxy(loglabelM, loglabelD):
    """
    -------------------------------------------------------------------
    On PC, verify that PC and proxy-drive content is the same.  
    This is called for both initial copies and content syncs.
    It asks to run mergeall and diffall separately: the former is
    fast, but the latter is slow for larger content collections.
    Similar code in verify*part2 couldn't be factored into this.
    Update [1.2]: now also run from export part 2 (PC) post syncs.
    -------------------------------------------------------------------
    """
    if VerifyProxy:
        userreply = input('\nVerify PC copy to proxy copy with mergeall (y or n)?')
        if userreply == 'y':
            print('Running mergeall')
            logto = join(LOGS, '%s--%s-verify-mergeall-log.txt' % (DATE, loglabelM))
            runpy(join(MALL, 'mergeall.py'), 
                  join(FROM, STUFF), join(TO, STUFF), '-report', '-skipcruft', '-quiet',
                  logpath=logto,
                  tailing=(9, 'Mergeall summary'))

        userreply = input('\nVerify PC copy to proxy copy with diffall (y or n)?')
        if userreply == 'y':
            print('Running diffall')
            logto = join(LOGS, '%s--%s-verify-diffall-log.txt' % (DATE, loglabelD))
            runpy(join(MALL, 'diffall.py'),
                  join(FROM, STUFF), join(TO, STUFF), '-skipcruft', '-quiet',
                  logpath=logto,
                  tailing=(7, 'Diffall summary'))    # 6=none, 7=1 unique (__bkp__)



def previewChanges(loglabel):
    """
    -------------------------------------------------------------------
    On PC, show PC~proxy deltas before saving or propagating, for sync
    runs only.  Like offerFixnamesRun(), this routes output to a file
    and scrapes its final report, to reduce output in the console.
    -------------------------------------------------------------------
    """
    if PreviewSyncs:
        userreply = input('\nPreview changes in the source tree (y or n)?')
        if userreply == 'y':
            print('Running preview of changes to be synced')
            logto = join(LOGS, '%s--%s-preview-log.txt' % (DATE, loglabel))

            runpy(join(MALL, 'mergeall.py'), 
                  join(FROM, STUFF), join(TO, STUFF), '-report', '-skipcruft', '-quiet',
                  logpath=logto,
                  tailing=(9, 'Preview summary'))

            userreply = input('Display preview\'s full output (y or n)?')
            if userreply == 'y':
                print(open(logto, 'r', encoding='utf8').read())
            
            userreply = input('Continue with sync (y or n)?')
            if userreply != 'y':
                print('Sync aborted.')
                shutdown()



def removeExistingOrStop(thedst, where):
    """
    -------------------------------------------------------------------
    In both initial-copy scripts, get the user's okay and then delete
    existing content trees on proxy or phone; shutdown if not approved.
    Also used in the phone-sync script before unzipping, in case a 
    DELTAS folder is present from a prior-sync abort or other source.
    deltas.py always removes an existing DELTAS folder before starting.
    All calls to this precede a zip-extract, which requires an empty dir. 

    Update [1.2]: all calls to this are also crucial, so required=True 
    for the new rmtree_FWP(): if the user okays the delete, the user 
    must rm on delete failures, else stop run to avoid content damage.  
    At last count, there were 5 calls to this, and 6 to rmtree_FWP().
    -------------------------------------------------------------------
    """
    if os.path.exists(thedst):
        print('Removing prior content on', where)
        userreply = input('Proceed with removal (y or n)?')
        if userreply != 'y':
            print('Run stopped for existing content.')
            shutdown()
        else:
            timefunc(lambda: rmtree_FWP(thedst, required=True))
            print('Starting unzip')



def moveZipToProxy(thezip, prxzip):
    """
    -------------------------------------------------------------------
    Run a copy+delete across filesystems.  This is used by PC scripts
    for both the initial copy's full zip, and the sync's deltas zip.
    prxzip is already a file in driveRootPath(); run early for messages.
    This cannot use a simple move, because from/to devices differ.
    -------------------------------------------------------------------
    """
    try:
        if os.path.exists(prxzip):
            os.remove(prxzip)
        shutil.copy2(thezip, prxzip)
        os.remove(thezip)

    except Exception as E:
        print('Zip move failed - Python exception:', E)
        print('Shutting down; check permissions and space and try again.')
        shutdown(bad=True)



def tail(filename, lines, message): 
    """
    -------------------------------------------------------------------
    A portable 'tail -n lines filename': show the last 'lines' lines
    in file 'filename'.  This is inefficient for large files as coded, 
    but seek-and-scan is complex, and memory use is not a concern here.
    -------------------------------------------------------------------
    """
    print(message + ':')
    lastfew = open(filename, 'r', encoding='utf8').readlines()[-lines:]
    for line in lastfew: 
        print(indent + line.rstrip())
    return lastfew   # tail lines for scraping



def runpy(*cmdargs, logpath=None, input=None, showcmd=False, tailing=None):
    """
    -------------------------------------------------------------------
    Run a Python command line, optionally sending stdout to a file
    named in logpath, and providing stdin text from bytes input.
    *cmdargs is individual words in the command line to run; the
    first is the pathname of the Python script to run.  This adds 
    the host Python's path to the front of *cmdargs automatically.

    tailing can be used iff logpath is a filename.  It is a tuple 
    (lines, message) whose items are passed to the tail() function
    along with the logpath name.  When used, the return value is 
    the result of tail(); else, the return value is elapsed time.
 
    Used on both PC and phone, portable to Win, Mac, Linux, Android.
    Caveat: subprocess.run() requires Python 3.5+, but this is from 6 
    years ago (2015), and anything older would be rare to find today.
    Caveat: can't use showRuntimes=ShowRuntimes arg; not yet defined.

    Update: now always adds the '-u' switch of Python (not script) to
    force unbuffered stdout/stderr.  Else, logfiles may be incomplete 
    if inspected while a long-running operation is in progress (e.g., 
    a shared-storage delete of a large __bkp__ folder on the phone).
    diffall has its own '-u' but it's used only for frozen apps/exes.

    [1.2] Check exit status of Python run and shutdown immediately if 
    the run failed (status != 0).  The Python exception will be printed
    to stderr (the console) too.  This should never happen... except 
    that it did, after a bad copy of a zipfile to app-private storage.
    -------------------------------------------------------------------
    """
    cmdargs = (sys.executable, '-u') + cmdargs
    if showcmd:
        print('Command:', cmdargs)

    start = time.perf_counter()
    if not logpath:
        res = subprocess.run(args=cmdargs, input=input)
    else:
        logfile = open(logpath, 'w', encoding='utf8')
        res = subprocess.run(args=cmdargs, stdout=logfile, input=input)
        logfile.close()

    if res.returncode != 0:
        print('Python run failed:', cmdargs)
        print('Exiting; please resolve and retry.')
        shutdown(bad=True)

    stop = time.perf_counter()
    if ShowRuntimes: 
        elapsed = stop - start
        print(indent + 'Runtime: %dm, %.3fs' % (elapsed // 60, elapsed % 60))

    if logpath and tailing:
        lines, message = tailing
        lastfew = tail(logpath, lines, message)
        return lastfew
    else:
        return elapsed



def timefunc(func):
    """
    -------------------------------------------------------------------
    Time any function call; could be a decorator, but also for lambdas.
    -------------------------------------------------------------------
    """
    start = time.perf_counter()
    func()
    stop = time.perf_counter()
    if ShowRuntimes: 
        elapsed = stop - start
        print(indent + 'Runtime: %dm, %.3fs' % (elapsed // 60, elapsed % 60))



# constants for next function
FileLimit = 260 - 1             # 259 in Py, including 3 for drive, N for UNC
DirLimit  = FileLimit - 12      # 247 in Py, after reserving 12 for 8.3 name



def FWP(pathname, force=False, limit=DirLimit, trace=False):
    r"""
    -------------------------------------------------------------------
    [1.2] Fix too-long paths on Windows (only) by prefixing as
    needed to invoke APIs that support extended-length paths.

    force=True is used to fix paths preemptively where needed,
    regardless of length.  Use for top-level paths passed to tools 
    that walk folders - including the shutil.rmtree() call here.

    The fix is a direct add for local-drive paths:
       'C:\folder...' => '\\?\C:\folder...'
   
    But requires a minor reformatting step for network-drive paths:
       '\\server\folder...' => '\\?\UNC\server\folder...'

    Either way, this prefix lifts the pathname length limit to 32k 
    chars, with 255 chars generally allowed per path component.
    This is true even on PCs not configured to allow long paths.

    Paths are also made absolute before prefixing per Windows rules, 
    so it's okay if pathname is relative as passed in; the absolute 
    and prefixed form returned here will be sent to file-op calls.
 
    For additional docs trimmed here, see Mergeall's fixlongpaths.py, 
    from which this was copied (learning-python.com/mergeall.html).
    -------------------------------------------------------------------
    """
    if not RunningOnWindows:
        # macOS, Linux, Android, etc.: no worries
        return pathname
    else:
        abspathname = os.path.abspath(pathname)       # use abs len (see above)
        if len(abspathname) <= limit and not force:   # rel path len is moot
            # Windows path within limits: ok
            return pathname
        else:
            # Windows path too long: fix it
            pathname = abspathname                    # to absolute, and / => \
            extralenprefix = '\\\\?\\'                # i.e., \\?\ (or r'\\?'+'\\')
            if not pathname.startswith('\\\\'):       # i.e., \\   (or r'\\')
                # local drives: C:\
                pathname = extralenprefix + pathname  # C:\dir => \\?\C:\dir
            else:
                # network drives: \\...               # \\dev  => \\?\UNC\dev
                pathname = extralenprefix + 'UNC' + pathname[1:]
            if trace: print('Extended path =>', pathname[:60])
            return pathname



def rmtree_FWP(folderpath, required=False):
    """
    -------------------------------------------------------------------
    [1.2] Call this everywhere, instead of shutil.rmtree() directly.  
    On Windows, the path must be prefixed with FWP() for longpaths  
    _before_ shutil.rmtree().  Per ahead, onerror doesn't suffice.
    This is just for folder removals in this package; the nested 
    Mergeall and ziptools do similar for all their file operations.  

    With this new scheme, longpath items in folders on Windows will 
    generally be deleted without messages.  As a last resort, ask the 
    user to remove the item if rmtreeworkaround() and its error handler
    fail, instead of reraising an exception that triggers a crash and
    confusing exception message.  Verify if required==True.

    We can ignore final clean-up deletes, but not initial content or
    deltas deletes - these both could lead to data-integrity issues.  
    Pass required=True for contexts that require a delete to continue.
    Deletes can fail for many reasons, including locks for files in use, 
    intentional read-only permissions, or system bugs (it happens).
    -------------------------------------------------------------------
    """
    try:
        fixfolderpath = FWP(folderpath, force=True)
        shutil.rmtree(fixfolderpath, onerror=rmtreeworkaround)

    except Exception as E:
        print('System error - cannot delete folder "%s"' % folderpath)
        print('\nException info:', E, end='\n\n')

        if  required:
            input('Please delete manually, and press enter/return here when finished...')
            if os.path.exists(FWP(folderpath)):
                print('Sorry - cannot continue with folder not empty; exiting.')
                shutdown(bad=True)
        else:
            print('Please delete manually at your discretion.')



def rmtreeworkaround(function, fullpath, exc_info):
    """
    -------------------------------------------------------------------
    Catch and try to recover from failures in Python's shutil.rmtree() 
    folder removal tool, which calls this function automatically on all
    system errors.  Raising an exception here makes shutil.rmtree() 
    raise too, and a return makes it continue with results that vary 
    per function (e.g., os.rmdir can recover, but os.scandir cannot).

    This is adapted from Mergeall, which documents it in full.  In short,
    deletions are not always atomic on Windows, which matters for folder
    contents; macOS may remove "._*' Apple-double files automatically 
    with their main file before they are reached here; and Windows may 
    similarly remove folders automatically with their associated files.
    Other failures may stem from permissions or in-use locks: propagate.

    Note: Windows auto-removal of folders with files might only occur in
    Explorer; a TBD, but if so, the test here is pointless but harmless.
    macOS auto-removal of Apple-double files is pervasive, and Windows 
    pending-delete loops (and longpaths) have been spotted in the wild.

    ----
    Update [1.2]: Windows auto-deletes

    The Windows auto-folder-deletion test has been disabled pending an
    unlikely use case.  Per recent testing, the auto-delete of a folder 
    with its similarly named file is not triggered by this system's file 
    ops.  It's also not just an Explorer thing, but a Windows concept 
    known as "Connected Files" which can be disabled in the registry.

    To test: save any web page completely on Windows, such that both 
    an HTML file and a similarly named folder with a "_files" suffix 
    (usually) are saved.  Then, deletion results vary by context:

    In Windows Explorer:
       Auto-deletes happen - deleting the file automatically deletes the 
       folder, and deleting the folder automatically deletes the file.  
       So it's not just folders: associated files can vanish too.

    In the shell (Command Prompt):
       Auto-deletes do NOT happen - deleting the file with "del" does not
       automatically delete the folder, and deleting the folder with 
       "rmdir /S" does not automatically delete the file.

    In Python (py -3):
       Auto-deletes do NOT happen - deleting the file with "os.remove" (a.k.a.
       "os.unlink") does not automatically delete the folder, and deleting the
       folder with "shutil.rmtree" does not automatically delete the file.

    Hence, this is an "Explorer thing" in the sense that its code or APIs
    must recognize Connected Files.  File ops used in the shell and Python,
    however, do not.  The latter makes this case moot in this system.  While 
    files and/or folders might be deleted outside this system's code while
    a delete is in progress here, that's a different issue, and out of scope.

    Nit: this assumes that website saves are representative of Connected Files
    in general.  Else, it's unclear how this system can handle both vanishing
    files and folders (errors are broad), but it very likely doesn't have to.

    See https://duckduckgo.com/?q=windows+Connected+Files" for more; since 
    this system won't invoke the auto-deletes, though, it's irrelevant here.
    Microsoft also covers its older but typically proprietary behavior here:
    https://learn.microsoft.com/en-us/windows/win32/shell/manage#connected-files

    ----
    Update [1.2]: code typo fixes

    Fixed two uncaught but rarely triggered typos:
    - RunningOnMac => RunningOnMacOS, which matters only on macOS
      when "._*" files are auto-deleted before reached here.
    - Import 'errno' else undefined, which matters only on Windows
      when folders trigger the pending-deletions loop here.

    ----
    Update [1.2]: Windows longpaths support, take 2

    This system now wraps folder paths in FWP() before every call to 
    shutil.rmtree, to avoid Windows path-too-long errors.  The FWP() 
    prefixing call is a no-op outside Windows; see its code above.

    An initial coding (see the defunct "take 1" below) tried calling 
    FWP() _here_ on errors only, but that doesn't work in all cases: 
    for folders, Python's shutil.rmtree first tries os.scandir() and
    issues a call here if it fails on a too-long-path error, but then 
    does nothing to recover but set a folder-contents list to empty. 

    Hence, there's no way to adjust here that allows shutil.rmtree 
    to continue, and the folder's removal simply fails on a not-empty 
    state.  Calling FWP() _before_ the call to shutil.rmtree--like 
    Mergeall does--avoids the issue, because os.scandir() will succeed.
    Code here must still handle other failures, including Windows 
    pending deletions and macOS AppleDouble auto-removals.

    While at it, the higher-level rmtree_FWP() removal function now 
    also asks the user to delete manually if all else fails, instead
    of crashing or risking content corruption.  All of which should 
    be _exceedingly_ rare, but stuff happens.

    For examples of the console messages and interaction of the new 
    scheme, as well as additional background info, see this file:
    _etc/examples/x-1.2-export-example/_etc/windows-fwp-fix-results.txt

    ----
    Update [1.2]: Windows longpaths support, take 1 (defunct)

    Try fixing longpaths on Windows on any error.  The FWP() function 
    lifts Window's 260ish-character limit for path names, and is 
    borrowed from Mergeall and ziptools.  Both of those systems 
    underly the deltas scripts, and already wrap all file calls in 
    FWP().  Hence, most ops are already longpath safe, but this must 
    be able to delete folders which ziptools makes.

    This is an issue here only for folder-delete calls, so it's okay 
    to adjust in this error handler instead of at top-level calls.  
    This is also not an issue if Windows users enable longpaths via 
    the registry or Python installs, but this cannot be expected, and
    one crash has already been seen.  Still, the fix here will only
    kick in on PCs that otherwise fail for long paths.
    -------------------------------------------------------------------
    """

    # initial attempt to fix with FWP() here removed: see docstr [1.2]

    # Windows only, directory deletes only
    if RunningOnWindows and function == os.rmdir:

        """
        # assume removed by Windows, or other 
        # moot per above - disable till use case arises [1.2]
        if exc_info[0] == FileNotFoundError:
            msg = '...ignored FileNotFoundError for Windows dir'
            print(msg, fullpath)
            return                                   # folder deleted with file?: proceed                               
        """

        # wait for pending deletes of contents
        timeout = 0.001                              # nit: need to try iff ENOTEMPTY
        while timeout < 1.0:                         # 10 tries only, increasing delays
            print('...retrying rmdir: ', fullpath)   # set off, but not just for pruning
            try:
                os.rmdir(fullpath)                   # rerun the failed delete (with FWP!)
            except os.error as exc:
                if exc.errno == errno.ENOENT:        # no such file (not-empty=ENOTEMPTY) 
                    return                           # it's now gone: proceed with rmtree
                else:
                    time.sleep(timeout)              # wait for a fraction of second (.001=1 msec)
                    timeout *= 2                     # and try again, with longer delay
            else:
                return                               # it's now gone: proceed with rmtree

    # macOS only, ignore file-not-found for AppleDouble files
    if RunningOnMacOS and exc_info[0] == FileNotFoundError:
        itemname = os.path.basename(fullpath)
        if itemname.startswith('._'):
            # assume removed by macOS, or other
            print('...ignored FileNotFoundError for AppleDouble', fullpath)
            return

    raise  # all other cases, or wait loop end: reraise exception to kill rmtree caller



def driveRootPath(path):
    r"""
    -------------------------------------------------------------------
    Given a path, return its prefix which identifies the root of the 
    drive on which it resides.  This is used to ensure that zipfiles 
    are stored on the USB drive root, regardless of user TO settings.

    This keeps chopping off the path's tail item until a mount point
    is found, or empty or no more progress.  It may suffice to call 
    os.path.splitdrive() on Windows, but that's a no-op on POSIX.

    Examples:
    
    macOS:
    >>> driveRootPath('/Users/me/MY-STUFF/Code')
    '/'
    >>> driveRootPath('/Volumes/SAMSUNG_T5/fold3-vs-deltas')
    '/Volumes/SAMSUNG_T5'
    >>> driveRootPath('.')
    '/'
    # also '/' for 'folder/file'
    
    Windows:
    >>> driveRootPath(r'C:\Users\me\Desktop\MY-STUFF')
    'C:\\'
    >>> driveRootPath('D:\\test-ads\\z-fold-3.txt')
    'D:\\'
    >>> driveRootPath('.')
    'C:\\'
    # also 'C:\\' for '\Users\me', 'C:folder\file', 'folder\file'
    
    Linux:
    >>> driveRootPath('/home/me/Download/fold3-vs-deltas')
    '/'
    >>> driveRootPath('/media/me/SAMSUNG_T5/fold3-vs-deltas')
    '/media/me/SAMSUNG_T5'
    >>> driveRootPath('.') 
    '/'
    # also '/' for 'folder/file'
    -------------------------------------------------------------------
    """
    path = os.path.abspath(path)      # normalize, use \ on win
    next = path
    while next:
        if os.path.ismount(next):     # found the root path
            return next
        else:
            head, tail = os.path.split(next)
            if head == next:
                return path           # give up: don't loop
            else:
                next = head           # go to next path prefix
    return path                       # give up: no path left



[Home page] Books Code Blog Python Author Train Find ©M.Lutz