Does Git warn me if a shorthand commit ID can refer to 2 different commits?

If cee157 can refer to 2 different commit IDs, such as

cee157eb799af829a9a0c42c0915f55cd29818d4 and cee1577fecf6fc5369a80bd6e926ac5f864a754b

will Git warn me if I type in git log cee157? (or Git 1.8.5.2 (Apple Git-48) allows me to type in git log cee1).

I think it should, although I can't find any authoritative source that says it would.


It should give you something like this:

$ git log cee157
error: short SHA1 cee157 is ambiguous.
error: short SHA1 cee157 is ambiguous.
fatal: ambiguous argument 'cee157': unknown revision or path not in the working tree.
Use '--' to separate paths from revisions, like this:
'git <command> [<revision>...] -- [<file>...]'

I just tested this on a real Git repository, by finding commits with duplicate prefixes like this:

git rev-list master | cut -c-4 | sort | uniq -c | sort -nr | head

This takes the list of revisions in master, cuts out the first 4 characters and throws away the rest, count the duplicates and sort numerically. In a my relatively small repository of ~1500 commits I found quite a few revisions with a common 4-digit prefix. I chose a 4-digit prefix because that seems to be the shortest legal length supported by Git. (Doesn't work with 3 digits or less, even if not ambiguous.)

Btw this was not a typo, I don't know why the error message about ambiguous SHA1 appears twice, regardless of the number of duplicate SHA1 (tried with 2 and 3):

error: short SHA1 cee157 is ambiguous.
error: short SHA1 cee157 is ambiguous.

(Both on stderr. Actually the entire output is on stderr, nothing on stdout.)

Tested in Windows:

$ git --version
git version 1.8.1.msysgit.1

I think it's safe to say that if your version is >= 1.8.1, Git will warn you of duplicates. (It will refuse to operate with duplicates.) I would guess that much older versions worked this way too.

UPDATE

When testing this, you need a minimum of 4-digit SHA1, because of int minimum_abbrev = 4 in environment.c. (Thanks @devnull for pointing that out!)


The original poster states:

I think it should, although I can't find any authoritative source that says it would.

The authoritative source can be found in the source code, get_short_sha1().

Quoting this:

if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
    return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);

and this:

if (!ds->candidate_checked)
    /*
     * If this is the only candidate, there is no point
     * calling the disambiguation hint callback.
     *
     * On the other hand, if the current candidate
     * replaced an earlier candidate that did _not_ pass
     * the disambiguation hint callback, then we do have
     * more than one objects that match the short name
     * given, so we should make sure this one matches;
     * otherwise, if we discovered this one and the one
     * that we previously discarded in the reverse order,
     * we would end up showing different results in the
     * same repository!
     */
    ds->candidate_ok = (!ds->disambiguate_fn_used ||
                        ds->fn(ds->candidate, ds->cb_data));

if (!ds->candidate_ok)
    return SHORT_NAME_AMBIGUOUS;

Moreover, tests also exist to ensure that the feature works as expected.