On Mon, Jul 12, 2010 at 8:36 PM, Nazri Ramliy <ayiehere@xxxxxxxxx> wrote:
> If I were to use a single "sed" invocation here's how it's going to look like:
>
> sed -e "s/^pick [0-9a-f]\+//" < "$1" > commit_headers.$$
>
> But doing so reminds me of this:
>
> On Sat, Jul 10, 2010 at 6:30 AM, Junio C Hamano <gitster@xxxxxxxxx> wrote:
>> Nazri Ramliy <ayiehere@xxxxxxxxx> writes:
>>
>>> + sed -i "s/^\([a-z]\+\) [0-9a-f]\+ /\1 $REPLACE_COMMIT_ID /" \
>>
>> This is not portable. Escaping an ERE element with a backslash does not
>> make it suitable for use in BRE that sed uses.
>
> Aren't we back to square one? Or am I missing something?
sed 's/^pick [0-9a-f]\{1,\} //'
is a valid posix BRE. Alternately + can be expressed as:
sed 's/^pick [0-9a-f][0-9a-f]* //'
>>> + diff "$EXPECT_HEADER" commit_headers.$$ > /dev/null
>>
>> Is "test_cmp" inappropriate here for some reason?
>
> It seems appropriate, but this will require a ". ./test-lib.sh" from inside
> lib-rebase.sh, which fails because at that point we are already in the trash
> directory for the test, the solution is to do a ". $TEST_DIRECTORY/test-lib.sh"
> instead but that gives more errors due to test-lib.sh assumes that it is always
> being "sourced" when $CWD is $TEST_DIRECTORY, i think.
Don't the scripts that source lib-rebase.sh all source test-lib.sh
ahead of it? lib-rebase.sh shouldn't need to re-source test-lib.sh.
No?
j.
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html