On 12/31/10 08:06, Eric Blake wrote:
> On 12/30/2010 06:37 PM, Bruce Korb wrote:
>> Hi,
>>
>> Is this fatal? If so, how come it continued?
>
> It's fatal to git-version-gen, which did not continue.
git-version-gen has but two fatal conditions: invalid arguments
yielding a usage message and an unreadable "tarball version file".
That is not this message, but might be clarified with:
v=`cat $tarball_version_file 2>&1` || {
echo "$0 error: unreadable tarball version file $1: $v" >&2
exit 1
}
In any event, the invocation is:
./git-version-gen .tarball-version
and the file ".tarball-version" does not exist, hence git-version-gen
should not fail at all. So, this message says, "fatal: ..."
and comes from git and all three "git" invocations redirect stderr to
/dev/null. The fact that we see it is a git bug. Error messages
should be directed to stderr and thus written to /dev/null.
So, git-version-gen is correct to continue, but git should fail
with a message that names the program that fails ("git") and
should direct the message to stderr.
Note to GIT list: the message in question:
fatal: ambiguous argument 'v0.1..HEAD': unknown revision \
or path not in the working tree.
Thanks! Regards, Bruce
_______________________________________________
Autoconf mailing list
Autoconf@xxxxxxx
http://lists.gnu.org/mailman/listinfo/autoconf
