In `string.gsub()`, the `%` character has special meaning and must be escaped to be treated verbatim, otherwise the "invalid use of '%' in replacement string" warning will show up.
This adds a verbatim() function for that purpose. It fixes this warning for situations where `'%` characters are in the current path (cwd), version control branch names, or in the previous `PROMPT` set by the user.
The below enables Cmder Fast Init mode for `cmd.exe` sessions. This is more like the Cmder 1.3.5 init process. See issue #1821
Cmder Fast Init mode bypasses or disables the following Cmder 1.3.6+ features:
* Git root and version detection. Defaults to `%cmder_root%\vendor\git-for-windows` if it exists.
* Path enhance validation before path modify so `%Path%` enhancements are forced.
* Recursive path add for `"%CMDER_ROOT%\bin"`
* Recursive path add for `"%CMDER_USER_BIN%\bin"` if `/c [user_config_folder` is specified.
* `/d` switch to enable debug output.
* `/v` switch to enable debug output.
Add `/f` to Cmder task as shown below t enable fast init:
_Note 1: This setting is invalid in Cmder `Powershell` and `Bash` sessions~_
_Note 2: Add `/t` also to see init timer output_
Cuts ~2.4 seconds off of init time.
If the git found is a shim, we need to follow the path to the actual git
directory to set the GIT_INSTALL_ROOT variable.
This happens when installing git via scoop for example.
## Rename user-profile.* user_profile.* to resolve#1806, #1675
* This is a backward compatible fix and will automatically and silently rename users '%cmder_root%/config/user-profile.\*' to '%cmder_root%/config/user_profile.\*' and '[user_specified_config_root]/user-profile.\*' to '[user_specified_config_root]/user_profile.\*' if the sources exist.
* Cmder.exe does this for cmd.exe sessions.
* The init scripts for bash and Powershell handles it for these shells
Fixes Powershell 5.1 error when Windows Software Restriction Policy is enabled:
bin\vendor\profile.ps1 : Cannot dot-source this command because it was defined in a different language mode. To invoke this command without importing its contents, omit the '.' operator.
At line:1 char:1
+ . 'bin\vendor\conemu-maximus5\..\profi ...
+ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ CategoryInfo : InvalidOperation: (:) [profile.ps1], NotSupportedException
+ FullyQualifiedErrorId : DotSourceNotSupported,profile.ps1
@MartiUK: Squashing to avoid adding multiple "fixed" commits
* add args to init.bat
* adding args to cmder launcher
* reworked command line parsing and added a /C [path] arg for individual user config location
* removed unnecessary includes
* make shell init scripts work with CMDER_USER_CONFIG
* update tasks and readme.md
* fix git version check
* readme updates
* add register/unregister back in
* fixed git version again
* removed error if user defines user-aliases store file was not present an init.bat launch
* added enhance_path method to only update path if required
* added enhance_path method to only update path if required
* fixed a path prepend issue in enhance path
* init.bat with args is executed outside cmder/conemu sets cmder_root properly. Thanks @DRSDavidSoft
* fixed enhance path append issue
* implements recursive `/bin` path enhancing.
https://github.com/cmderdev/cmder/issues/1624
* added max depth
* changes command line arg to max_depth
* set max_depth default
* readme.md updates
* add back cmder /c [path] arg so it can be used with admin sessions since the env is not shared.
* readme.md updates
* fix /c setting of cmder_user_config
* changelog and readme
* remove bad arg
* fixed command line parsing and updated command line help on error
* Fixed ConEmu.xml file handling so it works again
* Added default user config root location if '/c' is specified and next arg is not another arg
* Replaced the git version comparison mechanism with actual version comparing
* Add notes regarding the git comparison method
Added a simple mechanism to call and compare the user installed git version against the vendored git version, rather than checking if the `git.exe` executable is located in a `/bin` directory.
This fixes false warnings by actually testing if the user installed git is older, or more recent.
Added a small note regarding how the `init.bat` works, and where to put custom `git.exe` binaries as well as where to find the Minified Cmder.
Aliases such as `ls` may have been overridden in the user's profile script and the resulting behavior may be different from `Get-ChildItem`. For example binding `ls` to GNU `ls.exe` generates an error if there are no matching files.
* Fix Cmder not changing to CMDER_START
Fix to allow Cmder to change to different startup directory if already on the same drive as CMDER_START
possibly relevant: HOME != USERPROFILE
* make "Cmder Here" work in more cases
integrated suggestion from comments (test for trailing backslash to prevent doubling of trailing backslash)
integrated suggestions from #1456: switching ```cd /d``` over to ```pushd``` to get support for UNC-paths
* replaced 'cd /d' with 'pushd'
replaced two more instances of 'cd /d' with 'pushd' to support network paths.
Also added double quotes to the use of '%USERPROFILE%' to prevent issues with spaces.
I also noticed that 'pushd', in contrast to 'cd /d' does not require the trailing backslash for drive roots (on Win7). Therefore I removed the trailing backslash check.
* Don't assume / is needed at end
* Update init.bat
(and some comments on the user-profile.cmd file)
Inspired by the comments from https://github.com/cmderdev/cmder/issues/193 and my personal need to use pageant instead of OpenSSH authentication agents (which is more Window user-friendly), I have used this approach which works as expected.
Keeping the spirit of the current scripts, I left it disabled, and with some comments explaining what they all do.
If the Install-Module command is available, do not include the bundled
module path containing a deprecated version of PSGet. Recent powershell
versions include PowerShellGet as standard module installer.
Resolve#1318.
Currently the "history" command in Cmder prints commands without any index associated with any command, so you can copy/paste a command or use "!!" to execute a command from the history but, unlike BASH, you can't do "!<command_index>".
"cat" has a switch, "-n", which prints line numbers of the file being printed. This, in conjunction with "history_io=3" in Clink settings (to read/write history when editing a command") works just like BASH, i.e., you can use the index printed next to a command in the history to execute that command as "!<command_index>".
Currently the "history" command in Cmder prints commands without any index associated with any command, so you can copy/paste a command or use "!!" to execute a command from the history but, unlike BASH, you can't do "!<command_index>".
"cat" has a switch, "-n", which prints line numbers of the file being printed. This, in conjunction with "history_io=3" in Clink settings (to read/write history when editing a command") works just like BASH, i.e., you can use the index printed next to a command in the history to execute that command as "!<command_index>".
This defines the default prompt variables ($PrePrompt, $PostPrompt,
$CmderPrompt, and $Prompt) before running the custom scripts (user-profile.ps1
and profile.d/*.ps1), and then saves those prompt values after running the
custom scripts. This gives the custom scripts and opportunity to override the
prompt code blocks.
This now opens the possibility for profile.d/*.ps1 scripts to affect the
prompt, and also allows both user-profile.ps1 and profile.d/*.ps1 scripts to
overwrite the $Prompt script block and overwrite the prompt behavior entirely.
Posh-Git before this release does not export `Write-VcsStatus` thus
powershell's autoloading cannot find the reqired module for the function.
note that `get-module -listAvailable` can return an array of multiple
versions.
E.g., if ${CMDER_ROOT} is /c/Users/Foo Bar/cmder,
the following errors will occur:
bash: pushd: /c/Users/Foo: No such file or directory
bash: [: /c/Users/Foo: binary operator expected
bash: /c/Users/Foo: No such file or directory
clink.get_cwd() is returning a string which is differently encoded than what
clink.prompt.value expects. This results in garbled path names if the path
condains non-ASCII chars. The (arguable hacky) solution is to parse the old
prompt for the current directory (which breaks if the user sets a PROMPT env var
which is incompatible to the regex used here...).
Also parse out a environment name set by systems like virtualenv or conda: this
could be done more specifically by targeting each such system and using the
usually set environment variable but this would mean that we would have to do
that for each and every such system out there and that is probably not a sane
idea...
This keeps the PROMPT variable as is and changes the prompt to the cmder style
in the clink code.
This has two advantages:
* opening a cmd in a cmder session will now show the old prompt code instead of
a ugly raw prompt without the replacements. This led to ugly output when a
batch file echoed their content (e.g `conda build recipe/`).
* when a command rewrites the prompt (e.g. an activate in a virtualenv), these
command sometimes simply overwrites the PROMPT so that the cmder enhancements
were not anymore in place. Now we simply don't care and overwrite it with our
stuff in the clink part. This might mean that a user has to install a lua
script so that e.g. conda environments are visible on the prompt.
Add a pre and post function hook around the Cmder prompt.
Specify the cmder prompt as a function that could be replaced by a user.
Write a friendly message when the user profile template is created.
Create the user profile with cmder prompt hooks ready to use.
It was concerning to run any function with a specific name every prompt
with no guarantee it remains what it was initally created as.
Core functions have been explicitly called from their
namespace like Microsoft.PowerShell.Utility\Write-Host to try and prevent
clobbering.
User supplied functions are passed in as script blocks, created as the
session runs the profile script. By creating them as constants these
function names cannot be declared again for the duration of the process.
Since the prompt function already exists by this time, set the readOnly
flag so to re-declare the prompt requires the use of -force.
It is hoped these changes limit what could be the risk of any script
redefining functions that are called automatically without user intent or
input.
There were problems when a path contained a parenthesis like `C:\temp\test (test)\`
As a precaution, quote all variables when they are used in echo or set.
The idea is:
* if the users points as to a specific git, use that
* test if a git is in path and if yes, use that
* last, use our vendored git
We don't make any attempt to guess a different location, if a user wants their
own git install, they have to choose "add git to path".
Also check that we have a recent enough version of git (e.g. test for
<GIT>\cmd\git.exe)
4bfdf04d59 moved all clink config files
to vendor/ and therefore the files in config/ were not used and .history
was also moved to vendor/.
The new arrangement uses `--script` to load our own clink.lua file which in
turn as the first thing loads the original clink.lua file. The user defined
files are again in config/. As clink regenerates these files, if they are not
present, this also means that we don't need to include the settings file
in the git repo and therefore don't overwrite it on updates.
The new system works by setting "CMDER_START" if a cmder.exe gets a
startup path. The rest is then implemented in the profile.ps1 and init.bat.
The new system works as follows:
1.If a startup path is given, cmder.exe sets this as `CMDER_START` and the
startup scripts change to that directory.
2.If the usere sets a `CMDER_START` environment variable and this is not
overwritten by cmder.exe, this variable is taken as the startup
directory by the startup scripts.
3.If not such path is give, the conemu startup dirs is the winner, as the
startup scripts don't change anything.
If having a long line of input in histroy, which went on into the
second row and using <up> to get back to it, you got the effect
that the input was split over three lines: "text, empty, text",
but using the arrows do move to the front of the first line was
moving the cursor in the second (empty) row. You could change
the text, but you needed to do it "in the dark". No idea why
s/os.execute/is.popen/ fixes this, but it does.
Partly adresses https://github.com/cmderdev/cmder/issues/749
