I've recently started using EDE (via CEDET via ECB) in an attempt to get Emacs set up as a reasonable IDE for development on a largish C/C++/Python project.
Something that was a bit fiddly to get set up (as I'm an Emacs and Lisp novice) was rapid file finding. The solution I currently have is to set up a raw/blank project (one EDE doesn't need to know how to manage makefiles or builds for) like so in my .emacs file:
;; Enable EDE mode
(global-ede-mode 1)
;; EDE knows nothing of my project type, so use CSCope to inform it of the project
;; and its files
(setq ede-locate-setup-options
'(ede-locate-cscope
ede-locate-base))
;; This is probably a dubious shortcut allocation,
;; but it's what I'm using at the mo.
(global-set-key (kbd "C-f") 'ede-find-file)
;; Set up the project for EDE
(ede-cpp-root-project "LargeApp" :file "/workspace/me/LargeApp/SConstruct")
M-x ede-find-file used to not work at all, because EDE didn't know anything of the files in the project directory and sub-directories. Setting up the files in an EDE project would have taken ages (10000+ files), but luckily EDE understands CSCope output, so it was a matter of doing this in my project root directory using a little bash script (based on information linked from the main CScope project page):
#! /usr/bin/env bash
cd /workspace/me/LargeApp
find /workspace/me/LargeApp \
-type d -name '.git*' -prune -o \
-type d -name '.svn*' -prune -o \
-type d -name 'build' -prune -o \
-type d -name 'bin' -prune -o \
-type d -name 'lib' -prune -o \
-name '*.h' -print -o \
-name '*.hpp' -print -o \
-name '*.c' -print -o \
-name '*.cpp' -print -o \
-name '*.cc' -print -o \
-name '*.py' -print -o \
-name '*.lua' -print -o \
-name '*.xml' -print >./cscope.files
cscope -b -q -k
EDE then picks up the cscope.out in the project root folder, and kazam!, M-x ede-find-file works.
However, there are one major niggle I'm struggling to sort out:
I need to type the full, case sensitive file name.
This is less than ideal on a project this size, because you often only remember part of the filename, and not necessarily the case. It would be great if I could get it set up so I only have to type a case-insensitive substring of the filename, and then get an IDO buffer or similar I can tab through until I get the file I'm looking for.
I've resorted to using M-x cscope-find-this-file instead. It forces me to choose from likely filepath options in a buffer, but it does a reasonable job of getting me the right file quickly and isn't hampered by case-sensitivity. If I can get this hooked up into IDO, then it should do exactly what I want.
For Python, CScope/xcscope.el is good, but not perfect for navigating Python tags. So I've started using Elpy. M-x elpy-goto-definition works very well for Python files, and for the rest (C/C++) I'm using M-x cscope-find-this-symbol.
Related
Sorry I am trying to configure ctags with emacs but I am in trouble.
I compiled global with this configure:
./configure --with-universal-ctags=/usr/local/bin/ctags
and I executed make && make install.
Then I changed the default target in the file .globalrc from native to new-ctags.
Finally I executed ggtags-create-tags within emacs.
Unfortunately I got the error
‘gtags’ non-zero exit: gtags: execvp failed.
gtags: unexpected EOF.
Can anyone help me, thanks
Using universal ctags is as simple as:
Run over a project (-R is to walk the project recursively, and -e is to use Emacs-compatible syntax):
$ ctags -eR
Alternatively if you like to only include files with certain extensions, you can use -a (append, creates a file if doesn't exist) option with find utility, like:
$ find -name "*.cpp" -print -or -name "*.h" -print -or -name "*.hxx" -print -or -name "*.cxx" -print | xargs ctags -ea
Run M-x visit-tags-table in Emacs, and navigate to the created TAGS file.
So I have a find command as below which finds the libclntsh.so.* files in a directory instantclient.
find instantclient -type f -name "*libclntsh\.so\.[0-9]*\.[0-9]*"
This results in for e.g.,
instantclient/libclntsh.so.11.1
How do I now ln within instantclient directory, ln -s libclntsh.so.11.1 libclntsh.so all with a find command in combination with exec
I should mention here that I DO NOT want to cd into instantclient.
And this is for Alpine Linux.
Use the -execdir option. As per manual:
-execdir command {} ;
Like -exec, but the specified command is run from the subdirectory containing the matched file, which is not normally the directory in which you started find. This a much more secure method for invoking commands, as it avoids race conditions during resolution of the paths to the matched files.
So your command will be:
find instantclient -type f -name "*libclntsh\.so\.[0-9]*\.[0-9]*" -execdir ln -s {} libclntsh.so \;
EDIT:
Another solution
find instantclient -type f -name "*libclntsh\.so\.[0-9]*\.[0-9]*" | xargs -I {} sh -c 'ln -s $(basename {}) instantclient/libclntsh.so'
I am new to gtags, and have a question. I have a big project, such as android AOSP, I want gtags to parse some folders, how can I achieve it with gtags? I searched and got solution:
use -f option with gtags, it seems doesn't support folders
Is there any good idea that I can set the folders path and gtags only process those folders?
UPDATE: author of the question came up with a better solution in the comments. I'm adding it here so it's easier to find:
.. create tag file in the sub-directories I need, and add the directories
to GTAGSLIBPATH when loading the project,
My answer:
You can limit what gtags indexes by adding list of files/directories to skip keyword in ~/.globalrc or /etc/gtags.conf. Here's a sample gtags.conf file.
The problem is that often global/gtags packages don't install gtags.conf (at least it's not there in global-5.7.1-2 on ubuntu 12.04), so you'll need to either get it from global source distribution, or use someone else's gtags.conf as a reference. For instance here.
Something like this should work. Note that leading / means from the top of the tree. Without it gtags will skip matching entries anywhere in the tree.:
common:\
:skip=/skip-this-dir/,/lib/and-this/,/include/and-this-one-too/:
The -f option is premised on find(1). Please try the followings.
$ find folder1 folder2 folder3 -type f -print | gtags -f -
or
$ find folder1 folder2 folder3 -type f -print >gtags.files
$ gtags
This is my bash function to get rid of files and paths including 'dummy' and 'win':
function gtagsupdate {
find . -name "*.c" -o -name "*.cpp" -o -name "*.h" -o -name "*.hpp" | grep -v dummy | grep -v win | gtags -f -
}
So the typical way I would create a diff log/patch between two branches in clearcase would to simply create two views and do a typical unix diff. But I have to assume that there is a more clearcase way (and also a '1-liner').
so knowing how to get a list of all files that have been modified on a branch:
cleartool find . -type f -branch "brtype(<BRANCH_NAME>)" -print
and knowing how to get the diff formatted output for two separate files:
cleartool diff FILE FILE##/main/PARENT_BRANCH_PATH/LATEST
so does anyone see any issues with the following to get a diff for all files that have been changed in a branch?
cleartool find . -type f -branch "brtype(CHILD_BRANCH)" -exec 'cleartool diff -ser $CLEARCASE_PN `echo $CLEARCASE_XPN | sed "s/CHILD_BRANCH/LATEST/"` ' > diff.log
Any modifications and comments are greatly welcomed
thanks in advance!
update: any ideas on how to get this too be a unix unified diff would also be greatly appreciated.
update2: So I think I have my solution, thanks go to VonC for sending me in the right directions:
cleartool find . -type f -branch "brtype(CHILD_BRANCH)" -exec 'cleartool get -to $CLEARCASE_PN.prev `echo $CLEARCASE_XPN | sed "s/CHILD_BRANCH/LATEST/"`; diff -u $CLEARCASE_PN.prev $CLEARCASE_PN; rm -f $CLEARCASE_PN.prev' > CHILD_BRANCH.diff
the output seems to work, I can read the file in via kompare without complaints.
The idea is sound.
I would simply make sure the $CLEARCASE_PN and $CLEARCASE_XPN are used with double quotes around them, to take into account with potential spaces in the file path or file name (as illustrated in "How do I list ClearCase versions without the Fully-qualified version?").
cleartool find . -type f -branch "brtype(CHILD_BRANCH)" -exec 'cleartool diff -ser "$CLEARCASE_PN" `echo "$CLEARCASE_XPN" | sed "s/CHILD_BRANCH/LATEST/"` ' > diff.log
Using simple quotes for the -exec directive is a good idea, as explained in "CLEARCASE_XPN not parsed as variable in clearcase command".
However, cleartool diff, even with the -ser (-serial) option don't produce exactly an Unix unified diff format (or Unified Format for short).
The -diff(_format) option is the closest, as I mention in "How would you measure inserted / changed / removed code lines (LoC)?"
The -diff_format option causes both the headers and differences to be reported in the style of the UNIX and Linux diff utility, writing a list of the changes necessary to convert the first file being compared into the second file.
One idea would be to not use cleartool diff, but use directly diff, since it can access in a dynamic view the right version through the extended pathname of the elements found.
The OP ckcin's solution is close that what I suggested with cleartool get:
cleartool find . -type f -branch "brtype(CHILD_BRANCH)" -exec 'cleartool get -to $CLEARCASE_PN.prev `echo $CLEARCASE_XPN | sed "s/CHILD_BRANCH/LATEST/"`; diff -u $CLEARCASE_PN.prev $CLEARCASE_PN; rm -f $CLEARCASE_PN.prev' > CHILD_BRANCH.diff
the output seems to work, I can read the file in via kompare without complaints.
In multiple line, for readability:
cleartool find . -type f -branch "brtype(CHILD_BRANCH)"
-exec 'cleartool get -to $CLEARCASE_PN.prev
`echo $CLEARCASE_XPN | sed "s/CHILD_BRANCH/LATEST/"`;
diff -u $CLEARCASE_PN.prev $CLEARCASE_PN;
rm -f $CLEARCASE_PN.prev' > CHILD_BRANCH.diff
(Note that $CLEARCASE_XPN and $CLEARCASE_PN are set by the cleartool find commant, they're not variables you set yourself.)
Transferring the answer from VonC and einpoklum to Windows I came up with the following. Create a separate batch file, which I called diffClearCase.bat, this eases up the command line significantly. It creates a separate tree for all modified files, which I personally liked, but the file and folders can be deleted afterwards.
#echo off
SET PLAINFILE=%1
SET PLAINDIR=%~dp1
SET CLEARCASE_FILE=%2
SET BRANCH_NAME=%3
SET SOURCE_DRIVE=T:
SET TARGET_TEMP_DIR=D:
SET DIFF_TARGET_FILE=D:\allPatch.diff
call set BASE_FILE=%%CLEARCASE_FILE:%BRANCH_NAME%=LATEST%%
call set TARGET_FILE=%%PLAINFILE:%SOURCE_DRIVE%=%TARGET_TEMP_DIR%%%
call set TARGET_DIR=%%PLAINDIR:%SOURCE_DRIVE%=%TARGET_TEMP_DIR%%%
echo Diffing file %PLAINFILE%
IF NOT EXIST %TARGET_DIR% mkdir %TARGET_DIR%
cleartool get -to %TARGET_FILE% %BASE_FILE%
diff -u %TARGET_FILE% %PLAINFILE% >> %DIFF_TARGET_FILE%
rem del /F/Q %TARGET_FILE%
And then I created a second bat file which simply takes the branch name as argument. In our case this directory contains multiple VOBs, so I iterate over them and do this per VOB.
#echo off
SET BRANCHNAME=%1
SET DIFF_TARGET_FILE=D:\allPatch.diff
SET SOURCE_DRIVE=T:
SET DIFF_TOOL=D:\Data\Scripts\diffClearCase.bat
IF EXIST %DIFF_TARGET_FILE% DEL /Q %DIFF_TARGET_FILE%
for /D %%V in ("%SOURCE_DRIVE%\*") DO (
echo Checking VOB %%V
cd %%V
cleartool find %%V -type f -branch "brtype(%BRANCHNAME%)" -exec "%DIFF_TOOL% \"%%CLEARCASE_PN%%\" \"%%CLEARCASE_XPN%%\" %BRANCHNAME%"
)
I am having problems with the greps in Emacs.
a) grep doesnt seem to understand the .[ch] for searching .c and .h files. This is a default option provided by Emacs with the lgrep command. The example is searching for the word "global" in .c/.h files.
grep -i -nH "global" *.[ch]
grep: *.[ch]: No such file or directory
Grep exited abnormally with code 2 at Mon Feb 16 19:34:36
Is this format not valid?
b) Using rgrep I get the following error:
find . "(" -path "*/CVS" -o -path "*/.svn" -o -path "*/{arch}" -o -path "*/.hg" -o -path "*/_darcs" -o -path "*/.git" -o -path "*/.bzr" ")" -prune -o -type f "(" -iname "*.[ch]" ")" -print0 | xargs -0 -e grep -i -nH "global"
FIND: Wrong parameter format
Grep finished (matches found) at Mon Feb 16 19:37:10
I am using Emacs 22.3.1 on Windows XP with the GNU W32 Utils (grep, find, xargs etc.). Grep v2.5.3 and find v4.2.20.
What am I missing?
UPDATE:
Too bad one can't accept multiple answers...since the solution to my problems are spread out.
grep -i -nH "global" *.c *.h
This solves the first problem. Thanks luapyad!
(setq find-program "c:\\path\\to\\gnuw32\\find.exe")
emacs was indeed using the Windows find.exe. Forcing the gnu32 find fixed the second problem. Thanks scottfrazer.
However, I still like ack best.
I found out that using:
(setq find-program "\"C:/path/to/GnuWin32/bin/find.exe\"")
(setq grep-program "\"C:/path/to/GnuWin32/bin/grep.exe\"")
Works better in windows, since you could have a space laying around the path and will screw up eventually.
Notice I used the two programs in my .emacs file.
Hope it's of some help to some other programmer in need ;)
Well, there is always Ack and Ack.el
For a) it looks like there are simply no .c or .h files in the current directory.
For b) Windows is trying to use its own find instead of the one from the GNU W32 Utils. Try:
(setq find-program "c:\\path\\to\\gnuw32\\find.exe")
Adam Rosenfield comment is worth expanding into an answer:
grep -r --include=\*.[ch] --exclude=\*{CVS,.svn,arch} -i -nH
To make the example given in this question work, use this:
grep -i -nH --include=\*.[ch] "global" *
It is also helpful to set the variable grep-command providing defaults to M-x grep:
(setq grep-command "grep -i -nH --include=\*.[ch] ")
Also here are some other useful command line parameters to grep:
-n print the line number
-s suppress error messages
-r recursive
I think the general problem is the windows cmd "shell" behaves very differently to a unix shell in respect to filename expansion regexps and wildcards.
To answer your (a) above try using:
grep -i -nH "global" *.c *.h
(You will still get an "invalid argument" if no *.c's or *.h's exist).
Or you can use command line option --include=\*.[ch] to make windows grep do "proper" filename pattern matching (see grep --help for other options)
I usually just use M-x grep and alter the command line args when prompted if I need to. But I just tried running M-x lgrep and got the same thing as you. It simply means that no files match *.[ch] in the current directory. You can customize the default options to include -r and search recursively through child directories as well:
M-x customize-group RET grep RET
Search for lgrep in that buffer to find/edit the Grep Template.
As far as M-x rgrep goes, I suspect it has something to do with the Windows version of find not liking the default options. That command works fine for me on Linux. Search for rgrep in that same customize buffer and tweak those options until the Windows find is happy.
Sorry I can't be more help with the Windows options, but I'm not familiar with them.