bug: `git log --grep ... --invert-grep --author=...` negates / ignores --author

Given the following commits:

asottile@asottile-VirtualBox:/tmp$ git init test
Initialized empty Git repository in /tmp/test/.git/
asottile@asottile-VirtualBox:/tmp$ cd test/
asottile@asottile-VirtualBox:/tmp/test$
GIT_COMMITTER_EMAIL=foo@example.com GIT_AUTHOR_EMAIL=foo@example.com
git commit --allow-empty -m "foo"
[master (root-commit) c9df62b] foo
asottile@asottile-VirtualBox:/tmp/test$ git commit -m "blah" --allow-empty
[master 9e3ee9b] blah
asottile@asottile-VirtualBox:/tmp/test$ git log
commit 9e3ee9bc1adab2ae8eb1884a8f6237da18dfd27b
Author: Anthony Sottile <asottile@umich.edu>
Date:   Wed May 31 08:40:59 2017 -0700

    blah

commit c9df62b93298a247fcfbe24ed4282ccf95448f47
Author: Anthony Sottile <foo@example.com>
Date:   Wed May 31 08:40:49 2017 -0700

    foo
asottile@asottile-VirtualBox:/tmp/test$ git log --grep bar
--invert-grep --author=foo
commit 9e3ee9bc1adab2ae8eb1884a8f6237da18dfd27b
Author: Anthony Sottile <asottile@umich.edu>
Date:   Wed May 31 08:40:59 2017 -0700

    blah

commit c9df62b93298a247fcfbe24ed4282ccf95448f47
Author: Anthony Sottile <foo@example.com>
Date:   Wed May 31 08:40:49 2017 -0700

    foo
asottile@asottile-VirtualBox:/tmp/test$ git log --author=foocommit
c9df62b93298a247fcfbe24ed4282ccf95448f47
Author: Anthony Sottile <foo@example.com>
Date:   Wed May 31 08:40:49 2017 -0700

    foo

I expect the same output from the last two commands, but the
`--invert-grep` one seems to match _all_ the commits.

I can try and dig into this if I have time, just trying to get a count
using this as a workaround

git log --grep ... --invert-grep --format=%ce | grep ... | wc -l

Anthony

On Wed, May 31, 2017 at 5:45 PM, Anthony Sottile [off-list ref] wrote:

Given the following commits:

asottile@asottile-VirtualBox:/tmp$ git init test
Initialized empty Git repository in /tmp/test/.git/
asottile@asottile-VirtualBox:/tmp$ cd test/
asottile@asottile-VirtualBox:/tmp/test$
GIT_COMMITTER_EMAIL=foo@example.com GIT_AUTHOR_EMAIL=foo@example.com
git commit --allow-empty -m "foo"
[master (root-commit) c9df62b] foo
asottile@asottile-VirtualBox:/tmp/test$ git commit -m "blah" --allow-empty
[master 9e3ee9b] blah
asottile@asottile-VirtualBox:/tmp/test$ git log
commit 9e3ee9bc1adab2ae8eb1884a8f6237da18dfd27b
Author: Anthony Sottile <asottile@umich.edu>
Date:   Wed May 31 08:40:59 2017 -0700

    blah

commit c9df62b93298a247fcfbe24ed4282ccf95448f47
Author: Anthony Sottile <foo@example.com>
Date:   Wed May 31 08:40:49 2017 -0700

    foo
asottile@asottile-VirtualBox:/tmp/test$ git log --grep bar
--invert-grep --author=foo
commit 9e3ee9bc1adab2ae8eb1884a8f6237da18dfd27b
Author: Anthony Sottile <asottile@umich.edu>
Date:   Wed May 31 08:40:59 2017 -0700

    blah

commit c9df62b93298a247fcfbe24ed4282ccf95448f47
Author: Anthony Sottile <foo@example.com>
Date:   Wed May 31 08:40:49 2017 -0700

    foo
asottile@asottile-VirtualBox:/tmp/test$ git log --author=foocommit
c9df62b93298a247fcfbe24ed4282ccf95448f47
Author: Anthony Sottile <foo@example.com>
Date:   Wed May 31 08:40:49 2017 -0700

11 more quoted lines

I had to squint a bit to see what you were getting at here, which is
pretty simple: When you provide --invert-grep the --author filter is
completely discarded. This is a bug.

I.e. on git.git:

OK:

$ git log --grep=bar --author=Ævar --pretty=format:%an -100 origin/pu
|sort|uniq -c|sort -nr
5 Ævar Arnfjörð Bjarmason

$ git log --author=Ævar --pretty=format:%an -100 origin/pu |sort|uniq
-c|sort -nr
100 Ævar Arnfjörð Bjarmason

$ git log --grep=bar --invert-grep --author=Ævar --pretty=format:%an
-100 origin/pu |sort|uniq -c|sort -nr
     78 Junio C Hamano
     14 Jeff King
      2 Andreas Heiduk
      1 Sahil Dua
      1 Rikard Falkeborn
      1 Johannes Sixt
      1 Johannes Schindelin
      1 Ben Peart
      1 Ævar Arnfjörð Bjarmason

That last command should only find my commits, but instead --author is
discarded.

On Wed, May 31, 2017 at 08:08:54PM +0200, Ævar Arnfjörð Bjarmason wrote:

$ git log --grep=bar --author=Ævar --pretty=format:%an -100 origin/pu
|sort|uniq -c|sort -nr
5 Ævar Arnfjörð Bjarmason

$ git log --author=Ævar --pretty=format:%an -100 origin/pu |sort|uniq
-c|sort -nr
100 Ævar Arnfjörð Bjarmason

$ git log --grep=bar --invert-grep --author=Ævar --pretty=format:%an
-100 origin/pu |sort|uniq -c|sort -nr
     78 Junio C Hamano
     14 Jeff King
      2 Andreas Heiduk
      1 Sahil Dua
      1 Rikard Falkeborn
      1 Johannes Sixt
      1 Johannes Schindelin
      1 Ben Peart
      1 Ævar Arnfjörð Bjarmason

That last command should only find my commits, but instead --author is
discarded.

I would have thought that the last command wouldn't find _any_ of your
commits. Don't we consider --author a greppable pattern and invert it?

By itself:

  $ git log --author=Ævar --invert-grep --format=%an -100 origin/pu |
    sort | uniq -c | sort -rn
       79 Junio C Hamano
        8 Stefan Beller
        8 Jeff King
        1 Sahil Dua
        1 Rikard Falkeborn
        1 Johannes Sixt
        1 Johannes Schindelin
        1 Andreas Heiduk

So that makes sense from the "--author is invertable" world-view.

But I'm puzzled that you show up at all when --grep=bar is added (and
moreover, that we get _one_ commit from you, not the 5 found in your
initial command). That seems like a bug.

-Peff

On Wed, May 31, 2017 at 11:40 PM, Jeff King [off-list ref] wrote:

On Wed, May 31, 2017 at 08:08:54PM +0200, Ævar Arnfjörð Bjarmason wrote:

quoted

$ git log --grep=bar --author=Ævar --pretty=format:%an -100 origin/pu
|sort|uniq -c|sort -nr
5 Ævar Arnfjörð Bjarmason

$ git log --author=Ævar --pretty=format:%an -100 origin/pu |sort|uniq
-c|sort -nr
100 Ævar Arnfjörð Bjarmason

$ git log --grep=bar --invert-grep --author=Ævar --pretty=format:%an
-100 origin/pu |sort|uniq -c|sort -nr
     78 Junio C Hamano
     14 Jeff King
      2 Andreas Heiduk
      1 Sahil Dua
      1 Rikard Falkeborn
      1 Johannes Sixt
      1 Johannes Schindelin
      1 Ben Peart
      1 Ævar Arnfjörð Bjarmason

That last command should only find my commits, but instead --author is
discarded.

I would have thought that the last command wouldn't find _any_ of your
commits. Don't we consider --author a greppable pattern and invert it?

There's two unrelated things going on here AFAICT:

* Anthony expected --author to be inverted by --invert-grep, but this
is not how it's documented, it's *only* for inverting the --grep
pattern: "Limit the commits output to ones with log message that do
not match the pattern specified with --grep=<pattern>."

* The --author filter should be applied un-inverted, but isn't, it's
seemingly just ignored in the presence of --grep, actually mostly
ignored, this is bizarre:

$ diff -ru <(git log --grep=bar --invert-grep --pretty=format:%an -100
origin/pu |sort|uniq -c|sort -nr) <(git log --grep=bar --invert-grep
--pretty=format:%an -100 --author=WeMostlyIgnoreThisButNotReally
origin/pu |sort|uniq -c|sort -nr)

--- /dev/fd/63  2017-06-01 21:44:08.952583804 +0200
+++ /dev/fd/62  2017-06-01 21:44:08.952583804 +0200

@@ -1,6 +1,6 @@
-     66 Junio C Hamano
+     65 Junio C Hamano
      10 Jeff King
-      8 Stefan Beller
+      9 Stefan Beller
       5 Lars Schneider
       2 SZEDER Gábor
       1 Tyler Brazier

By itself:

  $ git log --author=Ævar --invert-grep --format=%an -100 origin/pu |
    sort | uniq -c | sort -rn
       79 Junio C Hamano
        8 Stefan Beller
        8 Jeff King
        1 Sahil Dua
        1 Rikard Falkeborn
        1 Johannes Sixt
        1 Johannes Schindelin
        1 Andreas Heiduk

So that makes sense from the "--author is invertable" world-view.

But I'm puzzled that you show up at all when --grep=bar is added (and
moreover, that we get _one_ commit from you, not the 5 found in your
initial command). That seems like a bug.

I think that's the only thing that's not a bug, i.e. we just find 1
match in the first 100 (note -100), sorry for the confusing example.

Anyway, much of the above may be incorrect, I haven't dug deeply
beyond just finding that something's funny going on and we definitely
have *some* bugs here.

I actually only expected the --grep to be inverted -- I think I'm on
the same page with what's documented.

I'd be happy to dig into the code and investigate this some more but I
am not familiar with the git codebase, any code hints on where to get
bootstrapped?

Anthony

On Thu, Jun 1, 2017 at 12:45 PM, Ævar Arnfjörð Bjarmason
[off-list ref] wrote:

On Wed, May 31, 2017 at 11:40 PM, Jeff King [off-list ref] wrote:

quoted

On Wed, May 31, 2017 at 08:08:54PM +0200, Ævar Arnfjörð Bjarmason wrote:

quoted

$ git log --grep=bar --author=Ævar --pretty=format:%an -100 origin/pu
|sort|uniq -c|sort -nr
5 Ævar Arnfjörð Bjarmason

$ git log --author=Ævar --pretty=format:%an -100 origin/pu |sort|uniq
-c|sort -nr
100 Ævar Arnfjörð Bjarmason

$ git log --grep=bar --invert-grep --author=Ævar --pretty=format:%an
-100 origin/pu |sort|uniq -c|sort -nr
     78 Junio C Hamano
     14 Jeff King
      2 Andreas Heiduk
      1 Sahil Dua
      1 Rikard Falkeborn
      1 Johannes Sixt
      1 Johannes Schindelin
      1 Ben Peart
      1 Ævar Arnfjörð Bjarmason

That last command should only find my commits, but instead --author is
discarded.

I would have thought that the last command wouldn't find _any_ of your
commits. Don't we consider --author a greppable pattern and invert it?

There's two unrelated things going on here AFAICT:

* Anthony expected --author to be inverted by --invert-grep, but this
is not how it's documented, it's *only* for inverting the --grep
pattern: "Limit the commits output to ones with log message that do
not match the pattern specified with --grep=<pattern>."

* The --author filter should be applied un-inverted, but isn't, it's
seemingly just ignored in the presence of --grep, actually mostly
ignored, this is bizarre:

42 more quoted lines