linux patch ignoring line numbers

问题:

I have an product (say version 1) and there are some custom changes made to it (say version 1.x). I am running a diff on these versions and creating a patch file.

Now there is a later version of the product (say version 2). I wish to apply the patch created from version 1 and version 1.x onto version 2. The difficulty here is that the line numbers will not remain constant. I understand there might be some failed hunks as well but atleast it will save me a lot of manual effort if I can apply the patch on version 2 (by ignoring the line numbers but by referring the lines below and above).

What's the best way to go about it? Any suggestions? Also any useful tools which can help here?

Thanks! Gaurav

回答1:

Patch will do that normally. You'll see it say something like

Hunk #1 succeeded at 6 (offset 5 lines).

when it applies a patch at a different line number from the original.

If some of the neighboring lines have changed, but there is still enough context to determine where to apply the patch, it will say with fuzz N where N is the number of context lines that didn't match.

回答2:

I was having the same problem with unified diffs. It worked using context diffs as shown bellow:

Creating the patch:

diff -c original.file modified.file > my.patch

Applying the patch:

patch -p0 --fuzz 500 --ignore-whitespace < my.patch

After that we need to check for mismatches:

1. Generate a diff on the destination folder between the original and the modified files

   diff -c destination.file.orig destination.file > second.patch
   

2. Compare the patch from the source folder with the patch from the destination folder. I generally do this using the midnight commander's compare files command on Linux+Mac and WinMerge on Windows.