* log --graph draws an orphaned branch with parents
@ 2026-10-13 19:00 herr.kaste
2026-11-03 19:25 ` herr.kaste
0 siblings, 1 reply; 5+ messages in thread
From: herr.kaste @ 2026-10-13 19:00 UTC (permalink / raw)
To: git
Hi! I have the following:
Checkout https://github.com/packagecontrol/thecrawl.git then draw the graph
$ git log --graph --all --oneline
a few pages down, (I don't have a more minimal reproduction, --all is required)
I see
What happened? (Actual behavior)
* | | 03b117815 Changed metadata of `Modelines`
* | | 30f679f1f Bulk edit
\ | |
* | | 5a2cecdd9 Initial revision
/ /
| | * 6062ecf4f (origin/feat/compare-registries) Add script to compare
two registries
| |/
| * 8f034813a Clarify fatal crawl skip logs
5a2cecdd9 is actually the birth of an orphaned branch, gitk also shows
it like that.
What did you expect to happen? (Expected behavior)
E.g. something like this:
● | | 03b117815 Changed metadata of `Modelines`
● | | 30f679f1f Bulk edit
● | | 5a2cecdd9 Initial revision
/ /
| | ● 6062ecf4f (origin/feat/compare-registries) ...
| |/
| ● 8f034813a Clarify fatal crawl skip logs
What's different between what you expected and what actually happened?
An obvious drawing error, 5a2cecdd9 is drawn as if it had parents, and
the rightmost lane goes to nirvana.
[System Info]
git version:
git version 2.56.0.windows.2
cpu: x86_64
built from commit: cc4dbf752a05efdc0e04e71fd3e8110d11bdd35c
sizeof-long: 4
sizeof-size_t: 8
shell-path: D:/git-sdk-64-build-installers/usr/bin/sh
rust: disabled
feature: fsmonitor--daemon
gettext: enabled
libcurl: 8.22.0
OpenSSL: OpenSSL 3.5.9 29 Sep 2026
zlib: 1.3.2
SHA-1: SHA1_DC
SHA-256: SHA256_BLK
default-ref-format: files
default-hash: sha1
uname: Windows 10.0 26200
compiler info: gnuc: 16.2
libc info: no libc information available
$SHELL (typically, interactive shell): C:\Dev\Git\usr\bin\bash.exe
Regards
Caspar Duregger
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: log --graph draws an orphaned branch with parents
2026-10-13 19:00 log --graph draws an orphaned branch with parents herr.kaste
@ 2026-11-03 19:25 ` herr.kaste
2026-11-03 21:27 ` Pablo Sabater
0 siblings, 1 reply; 5+ messages in thread
From: herr.kaste @ 2026-11-03 19:25 UTC (permalink / raw)
To: git; +Cc: Pablo Sabater, Karthik Nayak, Chandra Pratap
Finally had time for a repro (tested with git 2.56.0.windows.3):
sh
git init -q -b main repro && cd repro
export GIT_COMMITTER_DATE="@1700000000 +0000"
GIT_AUTHOR_DATE="@1700000000 +0000"
git commit -q --allow-empty -m B
git switch -q -c topic
git commit -q --allow-empty -m T
git switch -q --orphan orphan
git commit -q --allow-empty -m R
git merge -q --allow-unrelated-histories --no-ff -m M main
git log --graph --oneline --all
Actual output:
* 40a81a9 M
|\
\ |
* | 1d9120a R
/
| * 5a29a85 T
|/
* f1cd12c B
With --no-graph-indent:
* 40a81a9 M
|\
* | 1d9120a R
/
| * 5a29a85 T
|/
* f1cd12c B
Regards
Caspar
Am Di., 13. Okt. 2026 um 21:00 Uhr schrieb herr.kaste <herr.kaste@gmail.com>:
>
> Hi! I have the following:
>
> Checkout https://github.com/packagecontrol/thecrawl.git then draw the graph
>
> $ git log --graph --all --oneline
>
> a few pages down, (I don't have a more minimal reproduction, --all is required)
> I see
>
> What happened? (Actual behavior)
>
> * | | 03b117815 Changed metadata of `Modelines`
> * | | 30f679f1f Bulk edit
> \ | |
> * | | 5a2cecdd9 Initial revision
> / /
> | | * 6062ecf4f (origin/feat/compare-registries) Add script to compare
> two registries
> | |/
> | * 8f034813a Clarify fatal crawl skip logs
>
> 5a2cecdd9 is actually the birth of an orphaned branch, gitk also shows
> it like that.
>
> What did you expect to happen? (Expected behavior)
>
> E.g. something like this:
>
> ● | | 03b117815 Changed metadata of `Modelines`
> ● | | 30f679f1f Bulk edit
> ● | | 5a2cecdd9 Initial revision
> / /
> | | ● 6062ecf4f (origin/feat/compare-registries) ...
> | |/
> | ● 8f034813a Clarify fatal crawl skip logs
>
> What's different between what you expected and what actually happened?
> An obvious drawing error, 5a2cecdd9 is drawn as if it had parents, and
> the rightmost lane goes to nirvana.
>
> [System Info]
> git version:
> git version 2.56.0.windows.2
> cpu: x86_64
> built from commit: cc4dbf752a05efdc0e04e71fd3e8110d11bdd35c
> sizeof-long: 4
> sizeof-size_t: 8
> shell-path: D:/git-sdk-64-build-installers/usr/bin/sh
> rust: disabled
> feature: fsmonitor--daemon
> gettext: enabled
> libcurl: 8.22.0
> OpenSSL: OpenSSL 3.5.9 29 Sep 2026
> zlib: 1.3.2
> SHA-1: SHA1_DC
> SHA-256: SHA256_BLK
> default-ref-format: files
> default-hash: sha1
> uname: Windows 10.0 26200
> compiler info: gnuc: 16.2
> libc info: no libc information available
> $SHELL (typically, interactive shell): C:\Dev\Git\usr\bin\bash.exe
>
> Regards
> Caspar Duregger
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: log --graph draws an orphaned branch with parents
2026-11-03 19:25 ` herr.kaste
@ 2026-11-03 21:27 ` Pablo Sabater
2026-11-03 21:29 ` Pablo Sabater
0 siblings, 1 reply; 5+ messages in thread
From: Pablo Sabater @ 2026-11-03 21:27 UTC (permalink / raw)
To: herr.kaste, git; +Cc: Pablo Sabater, Karthik Nayak, Chandra Pratap
On Tue Nov 3, 2026 at 7:25 PM WET, herr.kaste wrote:
> Finally had time for a repro (tested with git 2.56.0.windows.3):
>
> sh
> git init -q -b main repro && cd repro
> export GIT_COMMITTER_DATE="@1700000000 +0000"
> GIT_AUTHOR_DATE="@1700000000 +0000"
> git commit -q --allow-empty -m B
> git switch -q -c topic
> git commit -q --allow-empty -m T
> git switch -q --orphan orphan
> git commit -q --allow-empty -m R
> git merge -q --allow-unrelated-histories --no-ff -m M main
> git log --graph --oneline --all
Hi!
Thanks for the repro and the bug report.
>
> Actual output:
The bug is because the indentation state GRAPH_PRE_ROOT is being
triggered, due to "R" being a root. What this state does is add a new
row to connect the indentation:
>
> * 40a81a9 M
> |\
> \ |
^^^
This one
> * | 1d9120a R
Then it pushes everything to the right.
> /
> | * 5a29a85 T
> |/
> * f1cd12c B
>
> With --no-graph-indent:
>
> * 40a81a9 M
> |\
> * | 1d9120a R
> /
> | * 5a29a85 T
> |/
> * f1cd12c B
The fix should be that the rendering of the graph is the same with the
indentation or without it, as there is nothing ambiguous in this case
that needs it.
I think I may be able to see how to fix it tomorrow, but I'm quite busy
until next week.
Regards,
Pablo
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: log --graph draws an orphaned branch with parents
2026-11-03 21:27 ` Pablo Sabater
@ 2026-11-03 21:29 ` Pablo Sabater
2026-11-03 21:51 ` herr.kaste
0 siblings, 1 reply; 5+ messages in thread
From: Pablo Sabater @ 2026-11-03 21:29 UTC (permalink / raw)
To: Pablo Sabater, herr.kaste, git; +Cc: Karthik Nayak, Chandra Pratap
On Tue Nov 3, 2026 at 9:27 PM WET, Pablo Sabater wrote:
>
> I think I may be able to see how to fix it tomorrow, but I'm quite busy
> until next week.
I forgot to ask if you want to try to fix it yourself, I'm happy to
review it :).
Pablo.
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: log --graph draws an orphaned branch with parents
2026-11-03 21:29 ` Pablo Sabater
@ 2026-11-03 21:51 ` herr.kaste
0 siblings, 0 replies; 5+ messages in thread
From: herr.kaste @ 2026-11-03 21:51 UTC (permalink / raw)
To: Pablo Sabater; +Cc: git, Karthik Nayak, Chandra Pratap
From 48bbcf965dc630251cc32f38d7efca5e6fb77c5e Mon Sep 17 00:00:00 2001
From: Caspar Duregger <caspar.duregger@gmail.com>
Date: Tue, 3 Nov 2026 22:38:22 +0100
Subject: [PATCH] graph: do not indent visual roots with columns to their right
Since a34c00d96 (graph: indent visual root in graph, 2026-07-14), a
visual root that has a child is drawn indented, connected to its child
by an extra GRAPH_PRE_ROOT line. When there are columns to the right
of the root, this breaks the graph:
* M
|\
\ |
* | R
/
| * T
|/
* B
graph_output_pre_root_line() writes the extra " " for the root column
without shifting the columns to its right, so they are drawn between
two columns. graph_output_commit_line() then pushes them one column
to the right, but the following GRAPH_COLLAPSING line is computed from
the mapping, which knows nothing about that offset. The column next
to R does not continue into the next line, and with more columns to
the right of R one of them goes nowhere.
But in this situation the indentation is not needed in the first
place. The columns to the right of the root collapse into its column
once it is rendered, so the collapsing line always leaves a blank
right below the root, and the next commit cannot end up vertically
adjacent to it.
Do not consider a commit a visual root when there are columns to its
right, which gives:
* M
|\
* | R
/
| * T
|/
* B
Signed-off-by: Caspar Duregger <caspar.duregger@gmail.com>
---
graph.c | 12 ++++++++++++
t/t4218-log-graph-indentation.sh | 18 ++++++++++++++++++
2 files changed, 30 insertions(+)
diff --git a/graph.c b/graph.c
index 28bef1b88..95eef3a95 100644
--- a/graph.c
+++ b/graph.c
@@ -969,10 +969,22 @@ static int graph_is_visual_root(struct git_graph *graph,
* already a column, so it will not be rendered below, the
* current commit has to act as the last commit and omit
* indentation.
+ *
+ * If there are columns to the right of the commit, they collapse
+ * into its column once it is rendered, so nothing can be rendered
+ * right below it and there is no need to indent it.
+ *
+ * * | <- visual root with a column to its right
+ * /
+ * | *
+ *
+ * Indenting it would also push these columns to the right, but
+ * the collapsing code does not know about that offset.
*/
return graph_is_visual_root_candidate(graph->commit, graph) &&
!(graph->commit_in_columns &&
graph->columns[graph->commit_index].is_merge_parent) &&
+ graph->commit_index + 1 >= graph->num_columns &&
flags->is_next_visible &&
(!flags->next_has_column || graph->visual_root_depth > 0);
}
diff --git a/t/t4218-log-graph-indentation.sh b/t/t4218-log-graph-indentation.sh
index 24dc9b497..b3b9f7c14 100755
--- a/t/t4218-log-graph-indentation.sh
+++ b/t/t4218-log-graph-indentation.sh
@@ -449,6 +449,24 @@ test_expect_success 'two unrelated merges' '
EOF
'
+test_expect_success 'visual root with columns to its right is not indented' '
+ create_orphan _70 && test_commit 70_A &&
+ git checkout -b _71 && test_commit 71_A &&
+ create_orphan _72 && test_commit 72_A &&
+ TREE=$(git write-tree) &&
+ MERGE=$(git commit-tree $TREE -p _72 -p _70 -m 72_merge) &&
+ git reset --hard $MERGE &&
+ lib_test_check_graph _72 _71 <<-\EOF
+ * 72_merge
+ |\
+ * | 72_A
+ /
+ | * 71_A
+ |/
+ * 70_A
+ EOF
+'
+
test_expect_success '--max-count treats the last visible commit as
the last commit' '
lib_test_check_graph --max-count=2 _8 _9 _10 <<-\EOF
* 10_A
--
2.56.0.350.g48bbcf965
Am Di., 3. Nov. 2026 um 22:29 Uhr schrieb Pablo Sabater
<pabloosabaterr@gmail.com>:
>
> On Tue Nov 3, 2026 at 9:27 PM WET, Pablo Sabater wrote:
> >
> > I think I may be able to see how to fix it tomorrow, but I'm quite busy
> > until next week.
>
> I forgot to ask if you want to try to fix it yourself, I'm happy to
> review it :).
>
> Pablo.
>
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2026-11-03 21:51 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-10-13 19:00 log --graph draws an orphaned branch with parents herr.kaste
2026-11-03 19:25 ` herr.kaste
2026-11-03 21:27 ` Pablo Sabater
2026-11-03 21:29 ` Pablo Sabater
2026-11-03 21:51 ` herr.kaste
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox