Re: Scripted clone generating an incomplete, unusable .git/config

[Jonathan Nieder <jrnieder@gmail.com>]

On Thu, Nov 11, 2010 at 12:55:48PM -0500, Daniel Barkalow wrote:
> On Thu, 11 Nov 2010, Jonathan Nieder wrote:

>> --- a/builtin/clone.c
>> +++ b/builtin/clone.c
>> @@ -667,6 +667,5 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
>>  	strbuf_release(&branch_top);
>>  	strbuf_release(&key);
>>  	strbuf_release(&value);
>> -	junk_pid = 0;
>>  	return err;
>>  }
>
> I believe that would cause it to remove the repository when it terminates, 
> regardless of whether it completed or not.

Ah, right, the second remove_junk() call is because of atexit().

So why does git clone keep running after the first remove_junk() call?
It seems that the signal is initially set up (by Python's popen()?) as
SIG_IGN.  I guess "git clone" should explicitly override that to be
SIG_DFL?

Here's a proof of concept.  It is not very good because it overrides
any previously set sigchain handlers (in case the "git" wrappers
start to use one) and because using SIG_DFL as a sigchain_fun feels
like violating an abstraction.

diff --git a/builtin/clone.c b/builtin/clone.c
index 19ed640..2f21a91 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -458,6 +458,7 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 	}
 	junk_git_dir = git_dir;
 	atexit(remove_junk);
+	sigchain_push_common(SIG_DFL);
 	sigchain_push_common(remove_junk_on_signal);
 
 	setenv(CONFIG_ENVIRONMENT, mkpath("%s/config", git_dir), 1);