| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|\
| |
| |
| |
| |
| | |
Conflicts:
doc/ranger.1
ranger/defaults/apps.py
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
This fixes the bug that some programs are killed even if they were run
with the "d" flag, if ranger is killed by closing its terminal with the
window manager instead of closed with :quit.
|
| | |
|
| | |
|
| |
| |
| |
| | |
https://bbs.archlinux.org/viewtopic.php?pid=1143920#p1143920
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| | |
rough explanation: before, the "t" flag would run the program in a
terminal by prepending something like "xterm -e " to the command.
If the command is "ls | less", it would result in "xterm -e ls | less".
This commit changes it so the result looks more like
"xterm -e sh -c 'ls | less'" and works as intended.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
https://github.com/hut/ranger/issues/45
hut:
I've been experiencing irregular segfaults in ranger-master for which
I can't figure out the reasons. Here is all the info I got:
- It started around the time of commit 5417dda
- I think it is a problem with curses' addstr/addnstr function
- It happens randomly, there is no specific action that causes it
- It happens more often with python3 than with python2
- It's most likely somewhere in here: git diff master 5417dda5^
ranger/gui/widgets/browsercolumn.py - but I don't see anything
wrong.
hut:
I nailed it down to 93601b1 and fixed it.
For what it's worth, this is how to reproduce it: (tested with
rxvt-unicode 9.15 and dwm but may work with any other tiling window
manager):
- Ensure that the setting display_tags_in_all_columns is set to True
(this is the default)
- Tag a file or directory with the "t" key
- Navigate so that this tagged file is the LAST file you see of a
column other the main column
- Open a new window (resulting in rangers window getting smaller by at
least one row)
- Close a window (resulting in rangers window growing back to the
original size)
- Watch rangers brain being splattered all over the sand
If you don't use a tiling window manager, you can do steps 1-3 and
then resize rangers window very quickly.
When resizing ranger manually, one of these messages is printed
instead of "Segmentation Fault":
- *** glibc detected *** python2.7: corrupted double-linked list: 0x0949cc98 ***
- python2.7: malloc.c:3964: _int_free: Assertion `nextchunk->fd_nextsize->bk_nextsize == nextchunk' failed.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
https://github.com/hut/ranger/issues/44#issuecomment-7558251
adam8157:
have 10 mkv files, four of them are "application/octet-stream"(and file
without options returns "EBML file, creator matroska"), others are
"video/x-matroska".
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
in python2, weak references are not equal to the original object when
tested with "==", and this breaks Tab._set_thisfile_from_signal and
Tab._on_tab_change in python2 if weak=True is used in signal bindings
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
this was requested here: https://github.com/hut/ranger/issues/36
adam8157:
Why non-text files' default opener is editor now? It invoked open_with
before.
I think the old way is better. We can't describe all types in
rifle.conf, so many types are defined as unknown non-text types, it's
not appropriate to open them all with editor.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| | |
I accidently wrote 'fm.thisfile.cf' which had to be changed to
'fm.thisfile'
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| | |
This is added to compensate for the fact that the history is not shared
anymore and you can't simply type gnHH to restore a closed tab.
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The Environment class was weird to begin with. It had attributes and
methods that belonged to other classes. For example, keybinding
management (the attributes keybuffer and keymaps) should go to gui.ui,
directory management (garbage_collect, get_directory) should be in
core.fm whereas entering directories (enter_dir) and managing history
(history_go) should be the task of a separate Tab class.
This commit fixes it, all references to env should be only for backwards
compatibility now. I still need to rewrite the tabbing API in
core.actions to work with the new system. Every tab that is opened will
have its own Tab instance with its own history and pointer. Tab, unlike
Environment, is no SignalDispatcher - Environment's signals were moved
to fm.
BEFORE: fm.env.cf
AFTER : fm.thisfile
BEFORE: fm.env.cwd
AFTER : fm.thisdir
BEFORE: fm.env.signal_bind("move", ...) # same for the "cd" signal
AFTER : fm.signal_bind("move", ...)
BEFORE: fm.env.keybuffer # same for fm.env.keymaps
AFTER : fm.ui.keybuffer
BEFORE: fm.env.get_directory("/usr/lib") # same for half of the env methods
AFTER : fm.get_directory("/usr/lib")
BEFORE: fm.env.get_selection() # same for the other half of the env methods
AFTER : fm.thistab.get_selection()
Old configurations and hacks should still work because I added a
compatibility layer for the Environment class which translates all
getters and setters to the respective call of the new class.
What will NOT work are the Environment signals, please bind them to fm
instead.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This is a result of https://github.com/hut/ranger/issues/43 :
adam8157:
Hmm... I found this behavior is different between running in Debian Sid
and Fedora 17.
Their python version are both 2.7.3, but when you touch a file named
foo.sh, and press "r", ranger in Fedora 17 displays bash as the first
opener, then editor and pager, but ranger in Debian Sid displays editor
first, then pager and bash.
[...]
hut:
Ok, with your help I found it. Actions.execute_file and
Actions.draw_possible_programs optimize the process by skipping the call
to file --mimetype -Lb and using the mime type that the python mimetypes
library found. The two methods return different mimetypes sometimes.
Since file is more reliable than the python mimetypes library, I'll
change it to use file always.
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
This reverts commit e4f37e51e91cc45b7660e3572a275c8dae6dbec9.
|