Re: Diffs from CVS keyword expansion

Anyone have a nice script for suppressing diffs caused by CVS keyword expansion when someone checked the kernel sources into CVS?

I have 200,000 lines of various forms of this mixed in with real changes

It is unclear who checked it into CVS, Microcross or Arch Linux. If it is Arch Linux finding the source for their 2.6.11.4-1 release would help a lot. I looked on their web site and couldn't locate it.

I'm starting understand why the support for MX21 on a CSB535FS has never been added to the mainline.

Jon Smirl writes:
 > Anyone have a nice script for suppressing diffs caused by CVS keyword
 > expansion when someone checked the kernel sources into CVS?

I usually just say (with GNU diff)

  diff -I $$ ...

which is crude, but should be enough to get rid of the hunks listed by you.

You might also want -x CVS if you have the metadata files.

On Fri, 1 Dec 2006, Juergen Ruehle wrote:

Also, if you are actually _using_ CVS to check it out, use

	cvs co -ko

I think, to not check out with keyword expansion at all.

(Also usable with "cvs diff -ko", I think)

		Linus

PS. Clueless user alert: I have happily not used CVS in years and years, 

On 12/1/06, Linus Torvalds <torvalds@osdl.org> wrote:

I don't have the original CVS, just a tarball snapshot with the keywords expanded.

I'm working with the Freescale iMX21 and the Linux port for it has never be added to the mainline kernel. I am extracting bits and pieces of it from various vendor tarballs.

Hello Jon,

Jon Smirl wrote:

I have a little script (called rmcvsid) in my private path, that just wraps a Perl oneliner:

	#! /bin/sh
	exec perl -i -p -e 's/\$(Id|Revision):.*?\$/\$$1: \$/' "$@";

This edits inplace all cvs keywords to be not expanded. If you call it for all files *before* you throw them into git, you don't need to suppress the diffs, because there are none.

You may have to add Source, Date and maybe others.

HTH Uwe

Uwe> #! /bin/sh Uwe> exec perl -i -p -e 's/\$(Id|Revision):.*?\$/\$$1: \$/' "$@";

Ow. My eyes hurt from that. How about we rewrite that as a native Perl script:

    #!/usr/bin/perl
    $^I = ""; # this is -i
    while (<>) {
      s/\$(Id|Revision):.*?\$/\$$1: \$/;
      print;
    }

Hi,

On Tue, 5 Dec 2006, Randal L. Schwartz wrote:

Hey, that's better! All of a sudden, I understand everything!

Not, Dscho

Hi,

Johannes Schindelin wrote:

I don't like that either, just because it's Perl :-)

I'm not fluent in Perl, and didn't know $^I. Actually, that's better, yes. (And my script is already rewritten.)

One thing I don't like about both scripts is, that the timestamp of a file changes with that, even if there are no changes. Anyone knows another predefined variable/trick that doesn't hurt performance (and readability) too much?

timtowtdi Uwe

Uwe> One thing I don't like about both scripts is, that the timestamp of Uwe> a file changes with that, even if there are no changes. Anyone knows Uwe> another predefined variable/trick that doesn't hurt performance (and Uwe> readability) too much?

You have to do two passes, or a pass-and-a-half, and do the rename/rewrite based on whether anything changed.

"Just a few more lines of code." :) If it's really important, I can be coaxed to dig it up.

merlyn@stonehenge.com (Randal L. Schwartz) writes:

That's kinda surprising coming from the resident Perl guru.

I am so used to the command line "-i -p" that I did not even know what $^I does, and I would have lost without your "# this is -i" comment. That alone makes the Uwe's one more readable for me ;-).

tisdag 05 december 2006 20:55 skrev Junio C Hamano:

May I suggest a compromise? : 

#!/usr/bin/perl -i -p s/\$(Id|Revision):.*?\$/\$$1: \$/;