[RFC] [pre-commit] Add check-new-file-license

[RFC] [pre-commit] Add check-new-file-license

[Tom de Vries]

I forgot to add a copyright notice to a file.

I came across gdb/contrib/license-check-new-files.sh, and ran it out of
curiosity, and it pointed out the file.

Add a pre-commit check check-new-file-license that runs the script.

The check is only run for new files, because it's fairly slow, about 2 seconds
for an empty file:
...
$ touch gdb/bla.c
$ git add gdb/bla.c
  ...
check-new-file-license...................................................Passed
- hook id: check-new-file-license
- duration: 2.88s

gdb/bla.c: None

   ...
[master 86d2adef7bc] try
 1 file changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 gdb/bla.c
...

I added support in gdb/contrib/license-check-new-files.sh to check the staging
area by representing it as a diff between HEAD and HEAD+STAGED.
---
 .pre-commit-config.yaml                | 10 ++++++++++
 gdb/contrib/license-check-new-files.sh | 16 ++++++++++------
 2 files changed, 20 insertions(+), 6 deletions(-)

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index fe8466433ac..1f1a217b887 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -112,6 +112,16 @@ repos:
       additional_dependencies: ["pyyaml"]
       always_run: true
       require_serial: true
+    - id: check-new-file-license
+      name: check-new-file-license
+      language: python
+      entry: gdb/contrib/license-check-new-files.sh
+      args: [-s, HEAD, HEAD+STAGED]
+      files: '^$'
+      additional_dependencies: ["GitPython", "scancode-toolkit"]
+      always_run: true
+      verbose: true
+      require_serial: true
   - repo: https://github.com/nmoroze/tclint
     rev: v0.8.0
     hooks:
diff --git a/gdb/contrib/license-check-new-files.sh b/gdb/contrib/license-check-new-files.sh
index b961780a23f..8e9c398db7a 100755
--- a/gdb/contrib/license-check-new-files.sh
+++ b/gdb/contrib/license-check-new-files.sh
@@ -110,14 +110,18 @@ def main(argv):
         print(f'not a git repository (or any parent up to mount point "{dir}")')
         sys.exit(2)
 
-    # Get from/to commits
-    fc = get_commit(repo, from_commit)
-    tc = get_commit(repo, to_commit)
+    if from_commit == "HEAD" and to_commit == "HEAD+STAGED":
+        diff = repo.index.diff("HEAD", R=True)
+    else:
+        # Get from/to commits.
+        fc = get_commit(repo, from_commit)
+        tc = get_commit(repo, to_commit)
+        diff = fc.diff(tc, paths=paths)
+        paths = [str(dir) for dir in args.paths.split(",")]
+        print(f'Scanning directories {",".join(f"{s}/" for s in paths)}...')
 
     # Loop over new files
-    paths = [str(dir) for dir in args.paths.split(",")]
-    print(f'Scanning directories {",".join(f"{s}/" for s in paths)}...')
-    for file in fc.diff(tc, paths=paths).iter_change_type("A"):
+    for file in diff.iter_change_type("A"):
         filename = file.a_path
         if not args.skip:
             print(f"checking licenses for {filename}... ", end="", flush=True)

base-commit: 17113bdf7c556b987c55b170acad223f24304298
-- 
2.51.0

Re: [RFC] [pre-commit] Add check-new-file-license

[Keith Seitz]

Hi,

On 6/5/26 7:10 AM, Tom de Vries wrote:
> I forgot to add a copyright notice to a file.
> 
> I came across gdb/contrib/license-check-new-files.sh, and ran it out of
> curiosity, and it pointed out the file.
> 
> Add a pre-commit check check-new-file-license that runs the script.

Having dealt with licensing issues for Fedora/RHEL, I whole-
heartedly support this move. It's one less thing for all of us
downstream maintainers have to do.

[If accepted, you may also add my Reviewed-By.]

Thank you!
Keith

Re: [RFC] [pre-commit] Add check-new-file-license

[Tom Tromey]

>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:

Tom> I forgot to add a copyright notice to a file.
Tom> I came across gdb/contrib/license-check-new-files.sh, and ran it out of
Tom> curiosity, and it pointed out the file.

I didn't know about this file.  Weird that it has a '.sh' extension.

Tom> Add a pre-commit check check-new-file-license that runs the script.

Tom> The check is only run for new files, because it's fairly slow, about 2 seconds
Tom> for an empty file:

Hmm that really is quite slow.  Is it 2 seconds per invocation or per
new file?

Also the script seems to run git but I wonder if that's even necessary
in this setup.

Tom

Re: [RFC] [pre-commit] Add check-new-file-license

[Tom de Vries]

On 6/17/26 7:45 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> I forgot to add a copyright notice to a file.
> Tom> I came across gdb/contrib/license-check-new-files.sh, and ran it out of
> Tom> curiosity, and it pointed out the file.
> 
> I didn't know about this file.  Weird that it has a '.sh' extension.
> 

Yeah.  Renaming it to .py show some pre-commit errors:
...
gdb/contrib/license-check-new-files.py:58:5: E722 do not use bare 'except'
gdb/contrib/license-check-new-files.py:106:9: E722 do not use bare 'except'
...

> Tom> Add a pre-commit check check-new-file-license that runs the script.
> 
> Tom> The check is only run for new files, because it's fairly slow, about 2 seconds
> Tom> for an empty file:
> 
> Hmm that really is quite slow.  Is it 2 seconds per invocation or per
> new file?
> 

I did an experiment with 10 empty files:
...
$ for n in $(seq 1 10); do f=gdb/bla-$n.c; touch $f; git add $f; done
$ pre-commit run check-new-file-license -v
check-new-file-license...................................................Passed
- hook id: check-new-file-license
- duration: 3.18s

gdb/bla-1.c: None
gdb/bla-10.c: None
gdb/bla-2.c: None
gdb/bla-3.c: None
gdb/bla-4.c: None
gdb/bla-5.c: None
gdb/bla-6.c: None
gdb/bla-7.c: None
gdb/bla-8.c: None
gdb/bla-9.c: None
...

So, its seems it's per invocation.  If there is a significant cost per 
file we could parallelize it, the current approach is serial.

An invocation without new files is fairly quick:
...
$ pre-commit run check-new-file-license -v
check-new-file-license...................................................Passed
- hook id: check-new-file-license
- duration: 0.24s
...

> Also the script seems to run git but I wonder if that's even necessary
> in this setup.

Git is used to detect the new files using a diff command.

Maybe running this directly rather than going through a git interfacing 
library would be faster:
...
$ git diff --name-only --cached --diff-filter=A
...

Thanks,
- Tom

Re: [RFC] [pre-commit] Add check-new-file-license

[Tom de Vries]

On 6/17/26 9:19 PM, Tom de Vries wrote:
>> I didn't know about this file.  Weird that it has a '.sh' extension.
>>
> 
> Yeah.  Renaming it to .py show some pre-commit errors:
> ...
> gdb/contrib/license-check-new-files.py:58:5: E722 do not use bare 'except'
> gdb/contrib/license-check-new-files.py:106:9: E722 do not use bare 'except'
> ...

I've submitted a fix ( 
https://sourceware.org/pipermail/gdb-patches/2026-June/228100.html ).

Thanks,
- Tom