The find-binary-files.sh QA script (which is based on "file -i") reports the following files as binary: sci-libs/punc/ChangeLog: application/octet-stream; charset=unknown sci-libs/punc/Manifest: application/octet-stream; charset=unknown sci-libs/punc/punc-1.5.ebuild: application/octet-stream; charset=unknown Whereas on my machine, file -i reports them as text/plain with utf-8 or us-ascii charset. The package in question hasn't been touched for almost one year, and hashes in the Manifest verify correctly here. Reporting this here because the problem persists since last Saturday (2014-02-22).
Probably it's because of new version of sys-apps/file that went stable recently(bug #501574) - not sure that infra is already update this. I detected one strange thing with new version of file while doing custom tests myself - it does not recognize dd dump of first 512 bytes of hdd(where MBR resides) as, well, MBR. But do it with dump of first 4096 bytes. Other tests proves no regression, nobody before me detects anything wrong with this particular versions, that was security stabilization, so i was decided to move on with stabling. However, all of that facts are just my suggestion about root of this situation - maybe problem cause is completely different.
(In reply to Sergey Popov from comment #1) > Probably it's because of new version of sys-apps/file that went stable > recently(bug #501574) - not sure that infra is already update this. Unlikely, tested with file-5.12, 5.15, and 5.17 here, and all of them correctly identify the files in question as text/plain.
The files on brambling were corrupted. I believe I fixed them? -A
(In reply to Alec Warner from comment #3) > The files on brambling were corrupted. I believe I fixed them? sci-libs/punc/punc-1.5.ebuild doesn't appear in the report any more, but sci-libs/punc/{ChangeLog,Manifest} are still there. Reopening.
Ok one more attempt ;)
Looks good now.