git am mangles commit author name.

Hi all,

Have a problem with git am. I am trying to apply a patch (created with
git format-patch) that I have in a file.
The file contains the line (along with the rest of the patch, of course):
From: username <emailaddress>

So I do the following: git am --signoff < patchfile.patch It applies just fine, but the commit author shows in in git log as: emailaddress <emailaddress> (i.e., the username is nowhere to be seen)

I'm using the latest git release, 1.7.1.1, on ubuntu intrepid.

Would anyone happen to have some suggestions as to what i'm missing, and how I can get git am to properly treat the commit author name?

Thank you all in advance, Daniel

Hi Daniel,

Daniel F wrote:

Because I am lazy I tried with the latest 1.7.2 release candidate. Here is what I did:

 $ git fetch git://repo.or.cz/git/jrn.git rr/svn-fe
 $ git checkout FETCH_HEAD
 $ git format-patch HEAD^..HEAD
 $ git reset --keep HEAD^
 $ git am --signoff <0001-*

The commit author showed up with name and email address as I expected. What did I do wrong?

Alternatively: could you send the first few lines of your .patch file?

Thanks, Jonathan

Hi Jonathan,

Thanks for your response!

On 7/13/10, Jonathan Nieder <jrnieder@gmail.com> wrote:

I figured it out :) This was the from line:
From: bc <email@somedomain.net>

Apparently, git-am fails when the username is just one word. if i put a space between the b and the c, it works just fine.

However... do you not think that it is still a bug? Is it not quite reasonable for a person to want to identify with just a nick, rather than a full first and last name?

And more to the immediate point... is there any way i can get git am to do what i want? :)

Thanks, Daniel

On Tue, Jul 13, 2010 at 10:54 AM, Daniel F <nanotube@gmail.com> wrote:

And is less than 3 characters.

This is happening due to the get_sane_name check in mailinfo.c. The rules for a "sane" name are actually from Linus' original code. From 2744b23 (Start of early patch applicator tools for git., 2005-04-11):

+static char *sanity_check(char *name, char *email)
+{
+       int len = strlen(name);
+       if (len < 3 || len > 60)
+               return email;
+       if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
+               return email;
+       return name;
+}

You could add an option to mailinfo to disable the sanity check and then plumb that option into its various callers.

j.

On 7/13/10, Jay Soffian <jaysoffian@gmail.com> wrote:

Ah yes, indeed... Well, is there any reason to even have that length sanity check in the first place? If someone wants to be identified with a nick of 1 or 2 chars, what's wrong with that?

I have approximately zero familiarity with various git internals, so not sure if anything else depends on this length checkbeing there... But if not, if it is essentially arbitrary... why add complexity with extra options, instead of removing complexity by simply taking out the minimum length check?

At any rate... seeing as how it is hard-coded at the moment, I guess my immediate solution would be to discuss with my patch submitter about choosing a longer nick :)

Thanks, Daniel

On Tue, Jul 13, 2010 at 15:48, Daniel F <nanotube@gmail.com> wrote:

I don't know. The whole thing seems a bit silly. I changed it:

    -   if (name->len < 3 || 60 < name->len || strchr(name->buf, '@')
||
                   +   if (name->len < 1 || 60 < name->len ||
strchr(name->buf, '@') ||

And nothing in the test suite broke. Removing the `name->len <' clause completely broke one test in t5100-mailinfo.sh however.

But I don't understand that part of Git, maybe it really will break horribly somewhere with:

    Author: av <a@v.net>

On Tue, Jul 13, 2010 at 17:11, Jay Soffian <jaysoffian@gmail.com> wrote:

[..]

When I was a student a long time ago there was another (European) student there with a single-letter name. His legal name was 'Z'. It was printed like that in all official documents I ever saw, including bank accounts and loan papers, so it wasn't just some nickname he was using.

-Tor

On Wed, Jul 14, 2010 at 14:16, Tor Arntsen <tor@spacetec.no> wrote:

A lot of people also use their nicknames for their real name in Git, and plenty of people have 1-3 letter nicks.

On 7/14/10, Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

so... any git maintainers have thoughts on how this should be addressed, or whether this should be addressed at all? would simply setting minimum length to 1 have any negative side effects?

Hi Daniel,

Daniel F wrote: 

I am not a git maintainer, but I would suggest writing a patch to do that. Then you can try it and see, and send us a copy of the patch when reporting the results.

My guess is Linus wanted to make sure he had meaningful names in patches he applies[1]. There is no technical limit I know of that would lead one to forbid an author name longer than 60 characters, either. So the questions are:

 - did any code learn to rely on those limits later?
 - are any _people_ relying on those limits now?

I suspect the answer to question 2, at least for the lower bound, is no.

Hope that helps, Jonathan

[1] See v0.99.2~51^2~21 (Start of early patch applicator tools for git., 2005-04-11) for the original script.