submitting a patch for a directory

nglenn has asked for the wisdom of the Perl Monks concerning the following question:

Replies are listed 'Best First'.
Re: submitting a patch for a directory by daxim (Curate) on Aug 16, 2012 at 14:21 UTC
I was having trouble figuring out the correct patch command Generally, `patch -p0 < /tmp/the.patch` suffices. `diff -urd oldDir newDir > /tmp/the.patch head -n 1 /tmp/the.patch # directory should show as relative path, # e.g. oldDir/foo/bar/quux.pl cd oldDir/.. # move above the directory to patch # to ensure we are in the correct position mv newDir newDir.disabled # it must not interfere with the # following test application of the patch patch -p0 < /tmp/the.patch mv newDir.disabled newDir # restore` [download] submitting patches for whole distributions Learn git, this will be the single most useful tool for contributing patches to Perl projects. Is it normal, rather, to make a patch file for each changed file and submit all of them? No, that's abnormal. Your usage of diff -urd was customary, i.e. make one patch each containing the all changes for one feature or bugfix across all files.	[reply] [d/l]
Re^2: submitting a patch for a directory by nglenn (Beadle) on Aug 16, 2012 at 14:43 UTC
The module I contributed to didn't use git. Is CPAN being ported to a git repository?	[reply]
Re^3: submitting a patch for a directory by daxim (Curate) on Aug 16, 2012 at 14:49 UTC
https://github.com/gitpan	[reply]
Re^3: submitting a patch for a directory by jgamble (Pilgrim) on Aug 17, 2012 at 01:37 UTC
No, it is up to the individual module owner to decide which, if any, version control system he or she will use. (The gitpan link submitted by daxim is not "official" and should be viewed as something to fork from to start your repository, if you decide to use github). I submitted a patch last year to a CPAN member whose code resides on sourceforge, and I'm sure there are coders out there who don't use version control systems at all beyond what CPAN/BackPAN provide.	[reply]
Re^2: submitting a patch for a directory by Anonymous Monk on Aug 16, 2012 at 20:42 UTC
Not just git; any DVCS that can quickly be applied to an existing source tree will do. Try Bazaar, Mercurial. I'm personally using Bazaar and find it a joy to use. (Of course, if the project has a git/bzr/hg repository, that tool will be the most convenient.)	[reply]
Re: submitting a patch for a directory (mini diff/patch tutorial) by Anonymous Monk on Aug 16, 2012 at 16:26 UTC
The situation ( file/directory tree ) Read more... (382 Bytes) Foo is the original directory (CGI.pm-3.59) Foo-old is the backup pristine copy (CGI.pm-3.59-old) Foo-new is thing you've edited that you generate the patch from (CGI.pm-3.59-new) Creating a new patch ( Foo-new.patch, CGI.pm-3.59-new.patch ) `$ diff -Npurd Foo-old Foo-new > Foo-new.patch` What the switches mean -N => --new-file Treat absent files as empty. -p => --show-c-function Show which C function each change is in. -u => -U NUM --unified=NUM Output NUM (default 3) lines of unified context. -r => --recursive Recursively compare any subdirectories found. -d => --minimal Try hard to find a smaller set of changes. NOTE: If you're making a patch from a directory like CGI.pm-3.59, where you've used perl Makefile.PL && make, esp if using -N, you should clean up before making a patch, you should make realclean What a patch looks like Read more... (609 Bytes) Applying your new patch to Foo (CGI.pm-3.59), what the recipient of that patch would do, is first change working directory, then call patch `$ cd Foo` `$ patch -p1 -i../Foo-new.patch` Read more... (291 Bytes) -p1 means ignore the first path part, ignore the bold parts --- Foo-old/two.txt 1969-12-31 16:00:00.000000000 -0800 +++ Foo-new/two.txt 2012-08-16 08:19:25.500000000 -0700 It also means don't create a Foo-new directory :) If you used -p0 (ignore no path parts) the patching would fail in two ways, it wouldn't find one.txt, and it would create a Foo-new directory Read more... (789 Bytes) For best patching experience Always use relative paths for `diff -Npurd diffolddir newdir` Always use the p1 or p0 option, otherwise patch will ignore all directories/patch the wrong file To re-create my original situation, to create files and directories out of a patch, first produce the patch `diff -Npurd emptyDir tmpFoo > tmpFoo.patch` [download] The new patch tmpFoo.patch looks like Read more... (1065 Bytes) Then apply the patch in your new working directory tmp2Foo Read more... (1137 Bytes) Oh noes, a mistake! I forgot to include the -p1 switch, and no directories were created! Trying again with -p1 Read more... (533 Bytes) Hooray! Hmm, I'll try p0 in the same directory Read more... (785 Bytes) Whoa, that worked also and it didn't interfere :) but I don't really need that extra tmpFoo, now to get rid of it without rm -rfv -- from the same working directory -Reversing a patch (just add -R to the command you used to apply the patch) Read more... (563 Bytes) See also http://perldoc.perl.org/5.14.0/perlhack.html#Patching, http://perldoc.perl.org/5.10.0/perlhack.html#Patching	[reply] [d/l] [select]
Re: submitting a patch for a directory by aitap (Curate) on Aug 16, 2012 at 14:05 UTC
To apply this patch you need to `cd` to the directory to be patched and run: `patch -p1 < .../.../your/patch/file`. `-p1` flag means "ignore one directory name from the beginning of the file paths in the patch", so file to be patched will be found and the resulting file will have the same name. Sorry if my advice was wrong.	[reply] [d/l] [select]