Importing diffs

Hi.

is there any reason why git insists diffs be -p1 format ?

it makes it hard to import a diff generated by svn (-p0).

thanks!

Because that's the default for what the Linux kernel project uses? :)

I'm pretty sure that svn can output -p1 patches, I've done it in the past...

thanks,

greg k-h

This only applies to traditional diffs, not to git diffs.

---

Try this patch. Completely untested; if it works, tell Junio. :)

 Documentation/git-apply.txt |    6 +++++-
 apply.c                     |    8 ++++++--
 2 files changed, 11 insertions(+), 3 deletions(-)

2490f3129103a0eae9013eb2a6f564f4a7290fbd

index 51c7d47..75076b6 100644
@@ -10,7 +10,7 @@ SYNOPSIS
 --------
 [verse]
 'git-apply' [--stat] [--numstat] [--summary] [--check] [--index] [--apply]
-	  [--no-add] [--index-info] [--allow-binary-replacement] [-z]
+	  [--no-add] [--index-info] [--allow-binary-replacement] [-z] [-pNUM]
 	  [<patch>...]
 
 DESCRIPTION

@@ -68,6 +68,10 @@ OPTIONS
 	backslash characters replaced with `\t`, `\n`, and `\\`,
 	respectively.
 
+-p<n>::
+	Remove <n> leading slashes from traditional diff paths. The
+	default is 1.
+
 --apply::
 	If you use any of the options marked ``Turns off
 	"apply"'' above, git-apply reads and outputs the

index c471a82..79e23a7 100644
@@ -19,6 +19,7 @@
 static const char *prefix;
 static int prefix_length = -1;
 
+static int p_value = 1;
 static int allow_binary_replacement = 0;
 static int check_index = 0;
 static int write_index = 0;

@@ -31,7 +32,7 @@ static int no_add = 0;
 static int show_index_info = 0;
 static int line_termination = '\n';
 static const char apply_usage[] =
-"git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [-z] <patch>...";
+"git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [-z] [-pNUM] <patch>...";
 
 /*
  * For "diff-stat" like behaviour, we keep track of the biggest change

@@ -217,7 +218,6 @@ static char * find_name(const char *line
  */
 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
 {
-	int p_value = 1;
 	char *name;
 
 	first += 4;	// skip "--- "

@@ -1799,6 +1799,10 @@ int main(int argc, char **argv)
 			excludes = x;
 			continue;
 		}
+		if (!strncmp(arg, "-p", 2)) {
+			p_value = atoi(arg + 2);
+			continue;
+		}
 		if (!strcmp(arg, "--no-add")) {
 			no_add = 1;
 			continue;

0.99.6.g3480

Also, be careful: the default for git-apply is very different from the default for a regular "patch".

"patch" without any "-p" at all will try to automagically figure out the right file, which has burnt me more than once when you have the same name (usually "Makefile") in multiple sub-directories and "patch" makes the wrong automagic guess.

git-apply with this patch will continue to use -p1. No guessing, no gray areas.

I do believe that the right thing to do is to just make SVN output "-p1" patches (I cannot imagine that you can't do so, since -p1 is a much saner format than -p0), but I guess teaching git-apply to take -pN for traditional patches is fine.

But if somebody suggests we do the automatic thing that "patch" does, I'll scream. Too many times have I been burnt by patch being "helpful" (in general, patch by default will try very hard to apply a patch, whether it makes sense or not).

Linus