Make version.dart ready for svn 1.7 checkouts

Make version.dart ready for svn 1.7 checkouts

As aprelev@ pointed out in http://dartbug.com/10193 subversion 1.7 changed the
way metadata is stored in svn checkouts.
Similar to git, svn 1.7 checkouts only have one '.svn' directory at the top
level.

Previously we had the assumption that all directories have a '.svn'.

R=ricow@google.com, whesse@google.com

Committed: https://code.google.com/p/dart/source/detail?r=22170

[Bill Hesse, 2013-04-29 11:58:56]

lgtm

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart
File tools/release/version.dart (right):

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart#ne...
tools/release/version.dart:197: var currentPath = new Path(new
Directory.current().path);
This is changing to Directory.current, so if version runs against the checked-in
binary, this will need to change when new ones are checked in.

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart#ne...
tools/release/version.dart:204: if (currentPath.toString() == "/") {
I'm not sure this will work on Windows - paths will start /C:/, and I guess
directoryPath will strip this down to /.  But there are also network shares:
//myshare/path/parts.

I don't have a really good solution for this, except path.append('..') might be
more likely to do the right thing.  But probably not.

Perhaps we could break one step earlier by putting currentPath =
currentPath.directoryPath before this test.  I think that would be slightly
safer, because we don't check '/'.

Also, I guess currentPath can't include .., but if it did, then using
.directoryPath would do the wrong thing, and go down a level, instead of up.

[ricow1, 2013-04-30 07:21:28]

LGTM

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart
File tools/release/version.dart (right):

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart#ne...
tools/release/version.dart:204: if (currentPath.toString() == "/") {
On 2013/04/29 11:58:56, Bill Hesse wrote:
> I'm not sure this will work on Windows - paths will start /C:/, and I guess
> directoryPath will strip this down to /.  But there are also network shares:
> //myshare/path/parts.
> 
> I don't have a really good solution for this, except path.append('..') might
be
> more likely to do the right thing.  But probably not.
> 
> Perhaps we could break one step earlier by putting currentPath =
> currentPath.directoryPath before this test.  I think that would be slightly
> safer, because we don't check '/'.
> 
> Also, I guess currentPath can't include .., but if it did, then using
> .directoryPath would do the wrong thing, and go down a level, instead of up.

I think we can fix this by doing:
do {


} while (currentPath != currentPath.directoryPath)

[kustermann, 2013-04-30 09:38:44]

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart
File tools/release/version.dart (right):

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart#ne...
tools/release/version.dart:197: var currentPath = new Path(new
Directory.current().path);
On 2013/04/29 11:58:56, Bill Hesse wrote:
> This is changing to Directory.current, so if version runs against the
checked-in
> binary, this will need to change when new ones are checked in.
Thanks for the hint.

https://codereview.chromium.org/14238039/diff/1/tools/release/version.dart#ne...
tools/release/version.dart:204: if (currentPath.toString() == "/") {
On 2013/04/30 07:21:28, ricow1 wrote:
> On 2013/04/29 11:58:56, Bill Hesse wrote:
> > I'm not sure this will work on Windows - paths will start /C:/, and I guess
> > directoryPath will strip this down to /.  But there are also network shares:
> > //myshare/path/parts.
> > 
> > I don't have a really good solution for this, except path.append('..') might
> be
> > more likely to do the right thing.  But probably not.
> > 
> > Perhaps we could break one step earlier by putting currentPath =
> > currentPath.directoryPath before this test.  I think that would be slightly
> > safer, because we don't check '/'.
> > 
> > Also, I guess currentPath can't include .., but if it did, then using
> > .directoryPath would do the wrong thing, and go down a level, instead of up.
> 
> I think we can fix this by doing:
> do {
> 
> 
> } while (currentPath != currentPath.directoryPath)

I think this is not doing what we want. It allows you to go through the loop
once if you invoke it with cwd '/' but not with cwd '/tmp'.

I changed it now to stop as soon as the root of the filesystem was scanned. (No
sane person has a checkout on '/' or 'C:\', but it's cleaner this way)

[kustermann, 2013-04-30 10:17:24]

Committed patchset #2 manually as r22170 (presubmit successful).