Re: git log anomalities

Uwe Kleine-König [off-list ref] writes:

Hello,

today I looked at the changes to drivers/net/ethernet/freescale/fec.c in
the kernel since v3.8 using

	git log --stat v3.8.. --full-diff -- drivers/net/ethernet/freescale/fec.c

which looks as expected. But when I added --graph the diffstats change.
E.g. for 793fc0964be1921f15a44be58b066f22b925d90b it changes from:

 drivers/net/ethernet/freescale/Makefile   |    3 +-
 drivers/net/ethernet/freescale/fec.c      | 1966 -----------------------------
 drivers/net/ethernet/freescale/fec_main.c | 1966 +++++++++++++++++++++++++++++
 drivers/net/ethernet/freescale/fec_ptp.c  |    3 -
 4 files changed, 1968 insertions(+), 1970 deletions(-)

to

|  Documentation/devicetree/bindings/net/dsa/dsa.txt |   91 +
|  .../bindings/net/marvell-orion-mdio.txt           |    3 +
|  Documentation/networking/ip-sysctl.txt            |   35 +-

[...]

|  404 files changed, 15373 insertions(+), 8563 deletions(-)

Is that a bug, or a feature I don't understand?

Nice catch.  It's a bad interaction between --full-diff, --stat and
--parents (which --graph implies in an internal-workings kind of way).

The parent rewriting gets to mess with the history *before* we generate
the diffs.  Normally this wouldn't matter, because the pathspec filter
would then again exclude all but the "real" change from the diff.  But
with --full-diff, you see all the changes between 793fc096 and the next
commit that touches the given pathspec (47a5247f), which is -- at this
stage, after rewriting -- the "parent".

I suspect to fix this we'll need to separate the "real" from the
"rewritten" parents, which might take a bit of work.

-- 
Thomas Rast
trast@{inf,student}.ethz.ch