Hi Ross,

On 10/27/23 17:29, Ross Burton via lists.openembedded.org wrote:
From: Ross Burton <[email protected]>

A previous patch[1] added the ability to allow the search pattern for
patches to be changed, so that patchreview can be used across the entire
meta-oe repository by changing the patterns.

However, this means the caller needs to write long patterns when calling
patchreview.

Instead, we can see if the specified directory contains a layer by
checking if conf/layer.conf exists.  If it does, then search for patches
inside this directory.  If it doesn't, assume that the specified
directory is a repository that contains sublayers (such as
meta-openembedded) and look through each of the directories that match
the pattern meta-*.

This means patchreview can both scan either a single layer (eg
.../poky/meta) or a repository of sublayers (eg .../meta-openembedded).

[1] oe-core 599046ea9302af0cf856d3fcd827f6a2be75b7e1

Signed-off-by: Ross Burton <[email protected]>
---
  scripts/contrib/patchreview.py | 36 +++++++++++++++++++++++++---------
  1 file changed, 27 insertions(+), 9 deletions(-)

diff --git a/scripts/contrib/patchreview.py b/scripts/contrib/patchreview.py
index 43de105adc2..af66e32e02e 100755
--- a/scripts/contrib/patchreview.py
+++ b/scripts/contrib/patchreview.py
@@ -41,7 +41,7 @@ def blame_patch(patch):
                                      "--format=%s (%aN <%aE>)",
                                      "--", patch)).decode("utf-8").splitlines()
-def patchreview(path, patches):
+def patchreview(patches):
      import re, os.path
# General pattern: start of line, optional whitespace, tag with optional
@@ -56,11 +56,10 @@ def patchreview(path, patches):
for patch in patches: - fullpath = os.path.join(path, patch)
          result = Result()
-        results[fullpath] = result
+        results[patch] = result
- content = open(fullpath, encoding='ascii', errors='ignore').read()
+        content = open(patch, encoding='ascii', errors='ignore').read()
# Find the Signed-off-by tag
          match = sob_re.search(content)
@@ -198,21 +197,40 @@ def histogram(results):
      for k in bars:
          print("%-20s %s (%d)" % (k.capitalize() if k else "No status", 
bars[k], counts[k]))
+def gather_patches(candidate):
+    # candidate can either be the path to a layer directly (eg meta-intel), or 
a
+    # repository that contains other layers (meta-arm). We can determine what 
by
+    # looking for a conf/layer.conf file. If that file exists then it's a 
layer,
+    # otherwise its a repository of layers and we can assume they're called
+    # meta-*.
+

Maybe mention that this expects a pathlib.Path object?

+    if (candidate / "conf" / "layer.conf").exists():
+        print(f"{candidate} is a layer")
+        scan = [candidate]
+    else:
+        print(f"{candidate} is not a layer, checking for sub-layers")
+        scan = [d for d in candidate.iterdir() if d.is_dir() and (d.name == "meta" or 
d.name.startswith("meta-"))]
+        print(f"Found layers {' '.join((d.name for d in scan))}")
+

What about just looking for all layer.conf files recursively instead?

e.g.:

scan = [l.parents[1].name for l in candidate.glob('**/conf/layer.conf')

This assumes a layer root directory is the parent of `conf` directory in which a layer.conf file is, which I think is a good assumption.

But this does not make any assumption that the layer should be prefixed with meta- or be meta, as opposed to the current implementation.

+    patches = []
+    for directory in scan:
+        filenames = subprocess.check_output(("git", "-C", directory, "ls-files", 
"recipes-*/**/*.patch", "recipes-*/**/*.diff"), universal_newlines=True).split()

FWIW, recipes- prefix is NOT enforced, it is derived from BBFILES variable in each layer.conf. (I have used layers where a typo was there, with recipe- or sometimes just recipes/). I am not saying we need to support those, but just merely raising this as a potential shortcoming.

Cheers,
Quentin
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#190071): 
https://lists.openembedded.org/g/openembedded-core/message/190071
Mute This Topic: https://lists.openembedded.org/mt/102223656/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to