On 10 Feb 2022, Stefan Sperling wrote:
On Thu, Feb 10, 2022 at 12:10:08AM -0600, Karl Fogel wrote:
I'm curious: what is the case that causes the patched target
file to have
different permissions than it had before the patch? (Or am I
misunderstanding what you're saying?)
Possibly related to the fact that the patched target is prepared
inside
a temporary file, which has 0600 permissions by default since it
gets
created by mkstemp(3) (see https://svn.apache.org/r880338).
This temp file is then renamed on top of the file in the working
copy and
file mode bits will be copied 1:1. We have had similar reports
before, where
files ended up being readable only by the user who was using the
working
copy, and where this interfered with some other use of such
files.
Ah, thank you, Stefan. That was the key point I didn't know
(until now).
Best regards,
-Karl
