MantisBT - Doomseeker
View Issue Details
0003262Doomseeker[All Projects] Bugpublic2017-09-12 03:332024-11-03 19:08
WubTheCaptain 
Zalewa 
lowtweakalways
closedfixed 
x86_64Debian GNU/Linuxbuster/sid
1.1 
1.4.01.4.0 
0003262: Allow $HOME and/or ~/ in file paths
Doomseeker currently requires the file path to be absolute. ~/ and $HOME are not recognized as valid alternatives to /home/<username>, instead /home/<username> (e.g. /home/wub in my case) is written to each configuration. In example for IWAD paths, ~/ and $HOME fail with "Iwad Path error".

This could possibly be extended to other user-configurable file paths, such as server executable path.

  1. Have a defined home folder for an Unix-like user. Mine is /home/wub. This is assumed to be the same as ~/ and $HOME.

  2. Have an IWAD of your choice in your preferred location somewhere in your home folder. Mine is /home/wub/.local/share/games/doom/freedoom.wad.

  3. Open the "Create Game" dialog in Doomseeker. If necessary, configure the minimum necessary to start a server.

  4. Attempt to start a server with IWAD path at preferred location: "/home/wub/.local/share/games/doom/freedoom.wad". This should work.

  5. Attempt to start a server with IWAD path at preferred location: "~/.local/share/games/doom/freedoom.wad". This should work, but fails to an Iwad Path error.

  6. Attempt to start a server with IWAD path at preferred location: "$HOME/.local/share/games/doom/freedoom.wad". This should work, but fails to an Iwad Path error.

Changing the path of my home folder (as a consequence of renaming my username on the local system) means all my previous configurations with IWAD paths in the "Create Game" window and various IWAD/PWAD paths subsequently need fixing to the new location.

If I recall correctly, at worst this means the "Create Game" window fails to list the additional WADs and they will be missing. The paths still exist in the .ini configuration and need to be manually edited to the new paths in a text editor.

In reverse: If I keep my old home location, no change is required if the user wants to hard-type their old path.
No tags attached.
related to 0003354closed Zalewa Portable mode is portable in name only. 
child of 0003279acknowledged  List of Debian issues (misc/non-policy) 
log dpkg.log (3,173) 2017-09-12 03:37
/tracker/file_download.php?file_id=2204&type=bug
Issue History
2017-09-12 03:33WubTheCaptainNew Issue
2017-09-12 03:37WubTheCaptainNote Added: 0018287
2017-09-12 03:37WubTheCaptainFile Added: dpkg.log
2017-09-12 03:38WubTheCaptainNote Edited: 0018287bug_revision_view_page.php?bugnote_id=18287#r10931
2017-09-21 10:40WubTheCaptainNote Added: 0018355
2017-09-21 10:43WubTheCaptainNote Edited: 0018355bug_revision_view_page.php?bugnote_id=18355#r10987
2017-09-27 18:30ZalewaRelationship addedchild of 0003246
2017-09-27 19:18WubTheCaptainNote Added: 0018398
2017-09-27 19:34WubTheCaptainNote Added: 0018399
2017-09-27 20:01ZalewaNote Added: 0018400
2017-09-27 21:53WubTheCaptainRelationship addedchild of 0003279
2017-09-27 21:54WubTheCaptainRelationship deletedchild of 0003246
2017-10-04 19:27WubTheCaptainCategorySuggestion => Bug
2017-10-05 02:44WubTheCaptainStatusnew => acknowledged
2017-12-11 16:49ZalewaRelationship addedrelated to 0003354
2018-08-27 03:44WubTheCaptainPriorityhigh => normal
2018-08-27 07:31ZalewaAssigned To => Zalewa
2018-08-27 07:31ZalewaStatusacknowledged => assigned
2018-09-04 11:59ZalewaNote Added: 0019489
2018-09-04 12:30ZalewaNote Added: 0019490
2018-10-05 06:17WubTheCaptainPrioritynormal => low
2018-10-05 06:17WubTheCaptainTarget Version => 1.3
2019-06-22 10:17ZalewaTarget Version1.3 =>
2022-09-10 14:59ZalewaTarget Version => 1.4.0
2022-10-26 20:49ZalewaNote Added: 0022476
2022-11-06 22:06ZalewaNote Added: 0022480
2022-11-09 21:42ZalewaNote Added: 0022484
2022-11-09 21:42ZalewaStatusassigned => feedback
2022-11-11 14:52WubTheCaptainNote Added: 0022488
2022-11-11 14:52WubTheCaptainStatusfeedback => assigned
2022-11-12 10:42ZalewaNote Added: 0022497
2022-11-27 14:30ZalewaNote Added: 0022501
2022-11-27 14:30ZalewaStatusassigned => needs review
2022-11-27 14:33ZalewaNote Edited: 0022501bug_revision_view_page.php?bugnote_id=22501#r13771
2022-11-29 22:36ZalewaNote Added: 0022502
2022-12-01 22:26ZalewaNote Added: 0022521
2022-12-01 22:27ZalewaNote Added: 0022522
2022-12-12 22:18ZalewaNote Added: 0022525
2023-01-05 12:13ZalewaNote Added: 0022643
2023-01-05 12:13ZalewaStatusneeds review => needs testing
2023-02-19 14:11ZalewaStatusneeds testing => resolved
2023-02-19 14:11ZalewaFixed in Version => 1.4.0
2023-02-19 14:11ZalewaResolutionopen => fixed
2024-11-03 19:08ZalewaStatusresolved => closed

Notes
(0018287)
WubTheCaptain   
2017-09-12 03:37   
(edited on: 2017-09-12 03:38)
Tested with Qt5 5.9.1 for original suggestion.

(0018355)
WubTheCaptain   
2017-09-21 10:40   
(edited on: 2017-09-21 10:43)
In addition to what's described in the steps to reproduce, here's some test case guidance for implementers (consider ~/ is /home/user1/):


  • ~user2/something.wad should fetch from /home/user2/something.wad, not /home/user1/user2/something.wad or /home/user1user2/something.wad.

  • ~/foo/~bar/something.wad should fetch from "/home/user1/foo/~bar/something.wad" (second tilde is escaped), not /home/bar/something.wad.

  • On Windows, for compatibility reasons a path beginning with ~ may optionally map to %HOMEDIR% (e.g. C:\Documents and Settings\user1 or C:\Users\user1 in example). Or, %HOMEDIR% could be supported too on Windows (if not already so).



QDir::homePath() may be useful.

(0018398)
WubTheCaptain   
2017-09-27 19:18   
After some consideration, I think this is an upstream issue with QFileDialog in Qt which requires no change in Doomseeker. The change isn't as trivial as I initially thought it would be. It'd be nice to have in Qt for sure.

Should we close this ticket?
(0018399)
WubTheCaptain   
2017-09-27 19:34   
Nevermind. This issue was fixed in Qt 4.8 six years ago:'https://bugreports.qt.io/browse/QTBUG-20571 [^]'

Seems like something about Doomseeker's use of QFileInfo is wrong, then. QFileInfo in Qt 5 definitely supports tilde.
(0018400)
Zalewa   
2017-09-27 20:01   
I'd still like to have a look around. Maybe I'll find out we're doing something non-standard with paths and it doesn't work everywhere. I'll also think if I want to support env vars resolution in paths.
(0019489)
Zalewa   
2018-09-04 11:59   
Progress update:

I've started working on this and already have a path resolver method that implements some placeholders - it's basic but it should be enough. Documentation excerpt looks like this:


/**
 * @brief Resolution mechanism for paths that can contain dynamically
 * resolvable placeholders.
 *
 * Allowed placeholders are:
 *
 * * `$PROGDIR` (capitals important) resolves to the directory where
 * the application's executable resides.
 * * `$` character followed by a letter or underscore followed optionally
 * by a string of letters, digits or underscores is treated as an env.
 * variable and resolution using such variable is attempted. If there's
 * no matching env. variable, the placeholder resolves to an empty string.
 * Regex: `$[a-zA-Z_][a-zA-Z0-9_]*`.
 * * `~` character at the beginning, followed by a directory delimiter
 * or end of the string is replaced with user's home directory.
 * The `~user` pattern is not allowed and will be treated as
 * a "~user" literal.
 *
 * While Windows platform quotes the env. variables with `%`, DataPaths
 * doesn't follow this pattern and uses the `$` prefix universally for
 * all platforms.
 *
 * @return Path with all placeholders resolved as best as possible.
 * The returned path is not normalized.
 */
QString resolveTemplated(const QString &templatedPath) const;


That was the easy part.

The more difficult part stems partially from the fact that a thing like this needs to be done with full intent from the very first moment of development of the program. Right now, finding all places where the resolution needs to happen as we hit the filesystem and all other places where we need to keep the templated path may prove to be a challenge. This problem isn't only limited to defining the configuration defaults, because user will expect that the placeholders can be used anywhere where path can be input. This is something that can be overcomed, however we need to be careful.

Another cause of problems is QStandardPaths. As much as it has proven to be helpful so far, it will now prove to be that much obstructive. Locations returned from it are absolute. Home dir is not represented as "~", but as an explicit path. To be even more general, Qt doesn't seem to have built-in support for "~" resolution, and anyway even if it had, we still desire to also use our own placehodlers for $PROGDIR and env. variables, which means we need custom resolver anyway. When we retrieve a path from QStandardPaths, do we need to check if it begins from home dir and replace this part with '~'? Where in the code should this happen? In DataPaths? Do we need our own wrapper over QStandardPaths?

Quote from "WubTheCaptain"
QFileInfo in Qt 5 definitely supports tilde.

Do you have any documentation on this? My tests indicate that this is not true.
(0019490)
Zalewa   
2018-09-04 12:30   
I've pushed the work done so far to a separate branch:'https://bitbucket.org/Doomseeker/doomseeker/commits/5d30d232dfaba9f8ce5bc07f6f79f65342a37a8e [^]'
(0022476)
Zalewa   
2022-10-26 20:49   
I decided to revisit this and hopefully complete it this time, as it still bothers me that the --portable mode isn't really portable when Windows assigns varying letter names to partitions depending on the computer you connect your portable USB drive to.

For now there's a commit that fixes $PROGDIRSOMETHING resolution bug and also implements handling of the "~user" pattern on UNIX-like systems:
'https://bitbucket.org/Doomseeker/doomseeker/commits/a1f315326a70aefc491415c0fe0f4d8345e06fc0 [^]'
plus some extra tests.
(0022480)
Zalewa   
2022-11-06 22:06   
The work continues... in the meanwhile:

The templated path resolution with $VARIABLE is getting complicated when someone really has a directory with a $ in its name. I can imagine someone naming their games directory "D:/game$/". There would need to be some escaping mechanism for this, such as $$ will take the single $ literally. And the paths retrieved from QFileDialog boxes would need to be translated to the escaped format, which means I will have to go through each case of where this is happening and deciding individually if the path needs to be escaped. Moreover, if the user sees "D:/game$$/" in their configured path instead of the "D:/game$/", they will think that the program has bugged their path, try to correct it manually and likely complain on #support.

This somewhat makes me think that the $VARIABLE resolution may not be the best idea. However, the overall work on this task can continue because the ~ resolution, while still affected by this problem, is less probable to get hit by it on accident.

I'm also identifying places where "absolutePath()" is used whereas just "path()" would do just fine.
(0022484)
Zalewa   
2022-11-09 21:42   
Continuing my monologue: I thought that maybe it would be sufficient and safe enough for the people if the $VARIABLE resolution was only limited if $VARIABLE was the first token in the path. We can limit the resolver to only resolve paths like "$DOOMDIR/wads", but not "/home/user/$DOOMDIRNAME/wads"?
(0022488)
WubTheCaptain   
2022-11-11 14:52   
All or nothing approach, imo. Fine with ~/ and ~user working as the first token in the path. $variable should be expandable anywhere, or the feature omitted if it cannot be implemented correctly at this time.
(0022497)
Zalewa   
2022-11-12 10:42   
Quote from WubTheCaptain
$variable should be expandable anywhere, or the feature omitted if it cannot be implemented correctly at this time.

I will have to consider how this will affect the 0003354, but yeah, not doing the $templates at all is an option on the table.
(0022501)
Zalewa   
2022-11-27 14:30   
(edited on: 2022-11-27 14:33)
I have pushed 4 more commits that should cover the remaining occurrences of where the paths should be resolved: 1, 2, 3, 4.

This isn't tested. I need to run the program now and verify it in runtime, so commits with fixes may still follow.

I also extracted the resolver to a separate Plugin API class and made it configurable a bit. The boolean switches can control if the ~, $PROGDIR, and $ENV_VAR should be resolved or not. For now this is all hardcoded to be enabled, but I'm strongly considering adding an user option to disable the $PROGDIR and $ENV_VAR resolution and have it disabled by default. This needs to be confronted with 0003354, which I will focus my attention at next.

(0022502)
Zalewa   
2022-11-29 22:36   
Option added to the "File Paths" config box:'https://bitbucket.org/Doomseeker/doomseeker/commits/8a219f6f7342128721ef611e5eb70714f578f0f4 [^]'

Also added one more unit test for $PROGDIR in the middle of the path:'https://bitbucket.org/Doomseeker/doomseeker/commits/180cd07cd08276b15a70310603237ee5c21dd971 [^]'
(0022521)
Zalewa   
2022-12-01 22:26   
Missed one conversion in the WADs picker.

Also it was necessary to add an explicit browse button to the file paths and the WADs picker, because "Add" needs to simply add an empty entry where the user can specify the path by hand. The browse box, which comes from the OS, and the templated paths don't mix well.
(0022522)
Zalewa   
2022-12-01 22:27   
The templated_paths branch is ready to be merged into master.
(0022525)
Zalewa   
2022-12-12 22:18   
Merged now:'https://bitbucket.org/Doomseeker/doomseeker/commits/55ed82145710397848538add0853df2b543e3182 [^]'
(0022643)
Zalewa   
2023-01-05 12:13   
Beta package for Windows available at the beta auto-update channel and at:'https://devbuilds.drdteam.org/doomseeker/doomseeker-1.3.3~beta-230105-1140_windows.zip [^]'

Please test. Keep in mind that the $ENVVAR and $PROGDIR placeholder replacements first need to be enabled in the "File paths" config box. Also keep in mind that the Windows build deliberately doesn't support the '~user' resolution (but it does support plain '~').