Re: [PATCH v5 1/1] cat-file: quote-format name in error when using -z

From: Junio C Hamano <gitster@pobox.com>
Date: May 15, 2023, 17:20 UTC
Message-ID: <xmqqmt25a4uk.fsf@gitster.g>
In-Reply-To: <ec139b78-1d36-f894-e39f-f29877a67b18@gmail.com>

Phillip Wood <phillip.wood123@gmail.com> writes:

Show 21 quoted lines

> On 12/05/2023 17:57, Junio C Hamano wrote:
>> Toon Claes <toon@iotcl.com> writes:
>> Stepping back a bit, how big a problem is this in real life?  It
>> certainly is possible to create a pathname with funny byte values in
>> it, and in some environments,letters like single-quote that are
>> considered cumbersome to handle by those who are used to CLI
>> programs may be commonplace.  But a path with newline?  Or any
>> control character for that matter?  And this is not even the primary
>> output from the program but is an error message for consumption by
>> humans, no?
>> I am wondering if it is simpler to just declare that the paths
>> output in error messages have certain bytes, probably all control
>> characters other than HT, replaced with a dot, and tell the users
>> not to rely on the pathnames being intact if they contain funny
>> bytes in them.
>
> We could only c-quote the name when it contains a control character
> other that HT. That way names containing double quotes and backslashes
> are unchanged but it will still be possible to parse the path from the
> error message. If we're going to munge the name we might as well use
> our standard quoting rather than some ad-hoc scheme.

In the above suggestion, I gave up and no longer aim to do "quoting". A more appropriate word for the approach is "redacting". The message essentially is: If you use truly problematic bytes in your path, they are redacted (so do not use them if it hurts).

This is because I am not sure how "names containing dq and bs are unchanged" can be done without ambiguity. If I see a message that comes out of this:

	printf("%s missing\n", obj_name);

and it looks like

	"a\nb" missing

how do I tell if it is complaining about the object the user named with a three-byte string (i.e. lowercase-A, newline, lowercase-B), or a six-byte string (i.e. dq, lowercase-A, bs, lowercase-N, lowercase-B, dq)?

If we were forbidding '"' to appear in a refname, then we could take advantage of the fact that the name of an object inside a tree at a funny path would not start with '"', to disambiguate. For the three- and six-byte string cases above, the formatting function will give these messages (referred to as "sample output" below):

	"master:a\nb" missing
	master:"a\nb" missing

because of your "we do not exactly do our standard c-quote; we exempt dq and bs from the bytes to be quoted" rule.

But it still feels a bit misleading. This codepath may have the whole objectname as a single string so that c-quoting the entire "<commit> <colon> <path>" inside a single c-quoted string that begins with a dq is easy, but not all codepaths are lucky and some may have to show <commit> and <path> separately, concatenated with <colon> at the outermost output layer, which means that the second one from the sample output may still mean the path with three-byte name in the tree of 'master' commit.

And worse yet, because

	git branch '"master'

is possible (even though nobody sane would do that), so "treat the string as c-quoted only if the object name as a whole begins with a dq", this disambiguation idea would not work. The first one from the sample output could be the blob at the path with a five-byte string name (i.e. lowercase-A, bs, lowercase-N, lowercase-B, dq) in the tree of the commit at the tip of branch with seven-byte string name (i.e. dq followed by 'master').

So, I dunno.