"make test" fails if /path/to/git.git contains spaces

I've noticed that "make test" fails if /path/to/git.git contains spaces, 
with the following error:

$ make test
make -C t/ all
make[1]: Entering directory `/home/Adam Roben/dev/git/t'
*** t0000-basic.sh ***
* error: cannot run git init -- have you built things yet?
make[1]: *** [t0000-basic.sh] Error 1
make[1]: Leaving directory `/home/Adam Roben/dev/git/t'
make: *** [test] Error 2

This is a very common configuration on Cygwin, but clearly not a common 
configuration for git developers, or else this error would have been 
corrected already.

Do we want to support having spaces in your path? It doesn't seem hard 
to fix, but it does seem like the kind of problem that will come up over 
and over again if only a very small set of people have this configuration.

-Adam

On Mon, Mar 31, 2008 at 5:51 PM, Adam Roben [off-list ref] wrote:

I've noticed that "make test" fails if /path/to/git.git contains spaces,
 with the following error:

 > $ make test
 > make -C t/ all
 > make[1]: Entering directory `/home/Adam Roben/dev/git/t'
 > *** t0000-basic.sh ***
 > * error: cannot run git init -- have you built things yet?
 > make[1]: *** [t0000-basic.sh] Error 1
 > make[1]: Leaving directory `/home/Adam Roben/dev/git/t'
 > make: *** [test] Error 2

 This is a very common configuration on Cygwin, but clearly not a common
 configuration for git developers, or else this error would have been
 corrected already.

 Do we want to support having spaces in your path? It doesn't seem hard
 to fix, but it does seem like the kind of problem that will come up over
 and over again if only a very small set of people have this configuration.

Turns out that while trying to fix the tests, I've found that
git-rebase doesn't like a $VISUAL having spaces, so perhaps this
should be tested more often :) Or should we just require $VISUAL have
no spaces?

I'll send a patch once I've got everything passing.

Thanks,

Bryan

Bryan Donlan schrieb:

On Mon, Mar 31, 2008 at 5:51 PM, Adam Roben [off-list ref] wrote:

quoted

 Do we want to support having spaces in your path? It doesn't seem hard
 to fix, but it does seem like the kind of problem that will come up over
 and over again if only a very small set of people have this configuration.

Turns out that while trying to fix the tests, I've found that
git-rebase doesn't like a $VISUAL having spaces, so perhaps this
should be tested more often :) Or should we just require $VISUAL have
no spaces?

I'll send a patch once I've got everything passing.

There was an attempt in the past to make this working. Please look at this
(rather lengthy) thread, in particular, my critique on the patches that
were submitted:

http://thread.gmane.org/gmane.comp.version-control.git/60544

I really would not like to see the same issues that I took the time to
comment on to show up again.

-- Hannes

On Tue, Apr 1, 2008 at 2:29 AM, Johannes Sixt [off-list ref] wrote:

Bryan Donlan schrieb:

quoted

On Mon, Mar 31, 2008 at 5:51 PM, Adam Roben [off-list ref] wrote:

quoted

quoted

 Do we want to support having spaces in your path? It doesn't seem hard

 >>  to fix, but it does seem like the kind of problem that will come up over
 >>  and over again if only a very small set of people have this configuration.
 >
 > Turns out that while trying to fix the tests, I've found that
 > git-rebase doesn't like a $VISUAL having spaces, so perhaps this
 > should be tested more often :) Or should we just require $VISUAL have
 > no spaces?
 >
 > I'll send a patch once I've got everything passing.

 There was an attempt in the past to make this working. Please look at this
 (rather lengthy) thread, in particular, my critique on the patches that
 were submitted:

 http://thread.gmane.org/gmane.comp.version-control.git/60544

 I really would not like to see the same issues that I took the time to
 comment on to show up again.

Thanks for the advice, I'll try not to make the same mistakes :)
I've found a few legitimate bugs as I've been updating the tests, so
I'd call this a worthwhile endeavour. I'll send the fixes in question
as broken-out patches in the final patch series.

Even if tests with space-ridden paths aren't run /often/, it'd still
be a good idea to fix them when they're noticed as broken, IMO. (I'm
currently testing with quotes and backslashes, too, now :)

Thanks,

Bryan