First of all, I apologize in advance for cc'ing the busy git
mailing list with problems about stgit.  The stgit web page at
http://www.procode.org/stgit/ says "Please report bugs to Catalin
Marinas and copy the GIT mailing list."  I would feel less guilty
about whose time I was spending on reports like this if there were a
separate stgit-users list.

	Anyhow, here is my bug reporting, which likely is some mistake
on my part.

	I have a git tree with snapshots of linux kernel tar files, no
branches and no local modifications.  By the way, this tree is not a
pull from the kernel.org git repository (although it will likely be
replaced with one in the future).  Rather, this git tree one of a
collection git trees that correspond to tar files released on various
ftp and web repositories (i.e., there is one for gcc, one for
binutils, etc.).

	I also have an stgit tree of the kernel.  This one has the
local modifications in stgit.  I just upgraded from stgit 0.11 to
0.12, upgraded the git kernel tree from 2.6.20 to 2.6.20-git7, and
tried to pull from the updated git tree to the stgit tree, using the
new stgit 0.12.  stgit refused to apply the first local patch,
complaining "stg pull: unknown user details."

	I did the "stg pull" as root, did not set any of the
GIT_AUTHOR_* environment variables and did not create any kind of git
or stg configuration file.  No doubt I could work around this by
setting up some kind of configuration file, but I should not be
required to do so, especially just to do a "pull".

	I don't mind that a version control system might log user ID
and email addresses somewhere, but it should still work if I do not
wish to provide that information.

Adam Richter
-

[ message continues ]

Hi Adam,


This is a known issue, which is reported in the StGit bugtracker as
https://gna.org/bugs/?8452
StGit should surely use the same info as git does.

Best regards,
-- 
Yann.
-

[ message continues ]

Gmail decided this mail is spam, probably because of the "apologize"

I asked the GIT people some time ago and they were OK with porcelain
tools postings here. They should let me know if it become annoying.
You could also use https://gna.org/bugs/?group=stgit for reporting

A quick fix is to run:

git repo-config --global user.name "Adam J. Richter"

As Yann said, there is already a bug logged for this issue. I also
think there is a different issue mailed me yesterday with the same
thing when pushing a patch, even though the patch author/committer
information was already in the patch metadata handled by StGIT.

I'll have a look at this.

-- 
Catalin
-

[ message continues ]

BTW, I wondered why we need to store author and committer info in the
patch dir.  Shouldn't the info from the Git commit be authoritative ?

Also, a quick glance at the code lets me think those information are
only changed in a patch when explicit flags are given to various
commands.  Shouldn't we, in addition set committer to current user
during refresh, and possibly non-forwarding push ?

Best regards,
-- 
Yann.
-

[ message continues ]

The initial idea was to store them only if the user requested on the
command line. I later added the configuration file and they were
always set.

I pushed a patch that allows stgit to work with the git defaults if
there is no configuration available. Let me know if you find it OK so
we can close the bug.

BTW, we should get a 0.12.1 release out with some bug-fixes.

-- 
Catalin
-

[ message continues ]

Right, fixing the pull bug should be done ASAP.  I'll try to have a
look at it tomorrow.
-

[ message continues ]