[PATCH] Make git-cvsexportcommit "status" each file in turn

Hi,

It turns out CVS doesn't always give the status output in the order
requested. According to my local CVS gurus this is a known CVS issue.

The attached patch just makes the script check each file in turn. It's
slower but correct.

I also slightly formatted the warn output when it detects problems as
multiple line wraps with long file paths where making my eyes bleed :-)

-- 
Alex, homepage: http://www.bennee.com/~alex/
Absence makes the heart go wander.

On Wed, Aug 15, 2007 at 02:27:28PM +0100, Alex Bennee wrote:

Hi,

It turns out CVS doesn't always give the status output in the order
requested. According to my local CVS gurus this is a known CVS issue.

The attached patch just makes the script check each file in turn. It's
slower but correct.

I also slightly formatted the warn output when it detects problems as
multiple line wraps with long file paths where making my eyes bleed :-)

I inlined the patch for easier commenting. Please inline further
patches.

This is extremly wastefull, because it will spawn a CVS process for each file.
A better fix would be to parse the filename from the output of
'cvs status' and use that as input for $cvsstat.

(And/or you could use an hash instead of an array for 'cvsoutput', so
you could double check that you only get the status for those files you
asked for.)

-Peter

On Wed, 2007-08-15 at 14:27 +0100, Alex Bennee wrote:

Hi,

<snip>

I also slightly formatted the warn output when it detects problems as
multiple line wraps with long file paths where making my eyes bleed :-)

Which I seem to have missed a bunch of crucial ;'s from

diff --git a/git-cvsexportcommit.perl b/git-cvsexportcommit.perl
index ee02c56..65c12b2 100755
--- a/git-cvsexportcommit.perl
+++ b/git-cvsexportcommit.perl

@@ -204,9 +204,9 @@ if (@canstatusfiles) {
 foreach my $f (@afiles) {
     if (defined ($cvsstat{$f}) and $cvsstat{$f} ne "Unknown") {
        $dirty = 1;
-       warn "File $f is already known in your CVS checkout.\n"
-       warn "  Perhaps it has been added by another user.\n"
-       warn "  Or this may indicate that it exists on a different

branch.\n"
+       warn "File $f is already known in your CVS checkout.\n";
+       warn "  Perhaps it has been added by another user.\n";
+       warn "  Or this may indicate that it exists on a different
branch.\n";
        warn "  If this is the case, use -f to force the merge.\n";
        warn "Status was: $cvsstat{$f}\n";
     }

-- 
Alex, homepage: http://www.bennee.com/~alex/
Love may laugh at locksmiths, but he has a profound respect for money
bags. -- Sidney Paternoster, "The Folly of the Wise"

On Wed, 2007-08-15 at 16:04 +0200, Peter Baumann wrote:

On Wed, Aug 15, 2007 at 02:27:28PM +0100, Alex Bennee wrote:

quoted

Hi,

It turns out CVS doesn't always give the status output in the order
requested. According to my local CVS gurus this is a known CVS issue.

<snip>
I inlined the patch for easier commenting. Please inline further
patches.

Will do. I assumed Evolution would do something sensible. My mistake :-(

quoted

---
 git-cvsexportcommit.perl |   30 ++++++++++++++++++++----------
 1 files changed, 20 insertions(+), 10 deletions(-)

diff --git a/git-cvsexportcommit.perl b/git-cvsexportcommit.perl

<snip>
This is extremly wastefull, because it will spawn a CVS process for each file.
A better fix would be to parse the filename from the output of
'cvs status' and use that as input for $cvsstat.

(And/or you could use an hash instead of an array for 'cvsoutput', so
you could double check that you only get the status for those files you
asked for.)

I agree it's wasteful and could be done better however I'm no perl
hacker so I just went for something that was correct and worked. 

The path that is echoed later in the status output is however the CVS
file path which may not be directly related to the actual path in your
source tree. For example I have one status reported as:

$ cvs status src/proj_version
===================================================================
File: proj_version      Status: Up-to-date

   Working revision:    1.1.380.1
   Repository revision: 1.1.380.1       /export/cvsroot/project/src/Attic/proj_version,v
   Sticky Tag:          ATAG (branch: 1.1.380)
   Sticky Date:         (none)
   Sticky Options:      (none)

This makes the matching more than a little problematic.

It depends on how much people that use this script care about performance?

For my part it's a fire and forget script once I've finished my hacking
in a git tree so I don't mind it taking some time. I'm not particularly
minded to dig further in perl to make it faster unless there is a real
clamour - or perhaps someone with a bigger itch and more perl foo can
tackle it. 

In the meantime it does fix a bug in the script so I would say it's
applying.

-- 
Alex, homepage: http://www.bennee.com/~alex/
Blessed is he who has reached the point of no return and knows it, for
he shall enjoy living. -- W. C. Bennett

onsdag 15 augusti 2007 skrev Alex Bennee:

On Wed, 2007-08-15 at 16:04 +0200, Peter Baumann wrote:

quoted

On Wed, Aug 15, 2007 at 02:27:28PM +0100, Alex Bennee wrote:

quoted

Hi,

It turns out CVS doesn't always give the status output in the order
requested. According to my local CVS gurus this is a known CVS issue.

<snip>
I inlined the patch for easier commenting. Please inline further
patches.

Will do. I assumed Evolution would do something sensible. My mistake :-(

quoted

quoted

---
 git-cvsexportcommit.perl |   30 ++++++++++++++++++++----------
 1 files changed, 20 insertions(+), 10 deletions(-)

diff --git a/git-cvsexportcommit.perl b/git-cvsexportcommit.perl

<snip>
This is extremly wastefull, because it will spawn a CVS process for each 

file.

quoted

A better fix would be to parse the filename from the output of
'cvs status' and use that as input for $cvsstat.

(And/or you could use an hash instead of an array for 'cvsoutput', so
you could double check that you only get the status for those files you
asked for.)

I agree it's wasteful and could be done better however I'm no perl
hacker so I just went for something that was correct and worked. 

The path that is echoed later in the status output is however the CVS
file path which may not be directly related to the actual path in your
source tree. For example I have one status reported as:

$ cvs status src/proj_version
===================================================================
File: proj_version      Status: Up-to-date

   Working revision:    1.1.380.1
   Repository revision: 

1.1.380.1       /export/cvsroot/project/src/Attic/proj_version,v

   Sticky Tag:          ATAG (branch: 1.1.380)
   Sticky Date:         (none)
   Sticky Options:      (none)

This makes the matching more than a little problematic.

CVS/Root contains the first part of the path. Just use the one found at the 
top of the CVS checkout. Using CVS may be insane, but mixing different repos 
in the same checkout is absolute madness.

Then drop /Attic/ if present and the ,v at the end of the file name. Not 
rocket science I'd say.

It depends on how much people that use this script care about performance?

We do although I rarely commit big patches so cvsexportcommit usually does 
it's job in seconds for me. I also use the -u option and don't do any work in 
the CVS checkout so performing status if just overhead for me. We might drop 
it completely with a switch.

For my part it's a fire and forget script once I've finished my hacking
in a git tree so I don't mind it taking some time. I'm not particularly
minded to dig further in perl to make it faster unless there is a real
clamour - or perhaps someone with a bigger itch and more perl foo can
tackle it. 

In the meantime it does fix a bug in the script so I would say it's
applying.

I suggest Junio wait a few days in case a real fix materializes.

-- robin