Re: [Bug report] git status doesn't escape paths of untracked files

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 10, 2020, 15:28 UTC
Message-ID: <xmqq363pr6r6.fsf@gitster.c.googlers.com>
In-Reply-To: <b48b2be0-c417-9140-54f1-73d0c59c40d5@web.de>

René Scharfe <l.s.r@web.de> writes:

Show 11 quoted lines

>>> +	struct strbuf sb = STRBUF_INIT;
>>> +	const char *rel = relative_path(in, prefix, &sb);
>>> +	int need_quotes = *rel != '"' && strchr(rel, ' ');
>>
>> relative_path() does not quote, so "begins with a dq" is not a good
>> test to see "if we were to pass this string to quote_c_style(), would
>> we get it back quoted already so we won't have to surround the
>> result with an extra pair of dq ourselves?".
>
> Ha!, that's true.  Makes me wonder how it was still able to pass the
> test suite, though..

The logic to refrain from ading extra dq-pair around the output from quote-c-style is "if it is already in a dq-pair, don't bother". The bug will trigger only when the input string has a character that needs to be quoted *and* SP in it. In such a case, relative_path() is likely to return *rel != '"' and you declare "we need to add a dq-pair ourselves", quote-c-style would give us a result in a dq-pair, and you add one extra layer.

But without a test with a path with both '"' and ' ' in it, such a bug would not trigger. Brian's test only checks what happens to a path with SP in it, without any other funny characters that need quoting with quote-c-style.