Re: calcsize locks /dev/null? Why?
Heiko Schlittermann <[email protected]> Sun, 24 Mar 2019 14:44:03 +0100
| Newsgroups | gmane.comp.archivers.amanda.devel |
|---|---|
| Organization | schlittermann -- internet & unix support |
| Message-ID | <[email protected]> |
Dustin J. Mitchell <[email protected]> (So 24 Mär 2019 14:32:17 CET): > The downside of testing with regard to locks is, depending on how the > timing works, an incorrect implementation may (almost) never fail. Yes, but here it seems to be quite clear that we are stuck on /dev/null, where dockerd has a shared lock. It seems, that the lock should ensure uninterrupted calcsize output to stderr. The mentioned githup patch seems to be ok too, as I suppose, that stderr is a channel to the parent process and no conflicts should arise with other applications. Using an extra lock file (as I did) isn't necessary probably. -- Heiko
signature.asc
(application/pgp-signature, 488 B)
-----BEGIN PGP SIGNATURE----- iQEzBAABCgAdFiEE0L/WueylaUpvFJ3Or0zGdqa2wUIFAlyXiaMACgkQr0zGdqa2 wUJWgwgAyqW1TlBvFhQ601Icy/uwyltnSv4OB0jpWhzUh+zrBHdVBp5zDFhTGowJ ysmTDh5aZ5nHYLLHFYirmyiP9Nk7WiJoedcu4nV0+QxOV8a4Yzmzw708MWhwkp8Q E2ToesJCf481FuLizPiYV6l1GcwOyVTAG6wZYZB8P47u2MUtr7BxX+SFUoRowAFE 59+/Vi/u+Sbgow788sDSve0gT0AbL4XOePtoTA+I1WxrY5NwvQXovRVu+5mWt2DA hC5wMH2OgtZNfbSejtsFl/hbUcUYFXgXU7LeMBgRyfSh7Fd+i6yldjoV/6NTg1Zs CbXduIi1+ix8UDOc5MMFdijPK0Vn5w== =jtWw -----END PGP SIGNATURE-----