On Sat, 8 Sep 2007 19:11:11 +0200 Herbert Laubner <[EMAIL PROTECTED]> wrote:
> See message below: > > Herbert Laubner <[EMAIL PROTECTED]> posted > [EMAIL PROTECTED], excerpted below, on > Sat, 08 > Sep 2007 10:14:01 +0200: > > > > I am installing xorg-x11 on an amd64 machine. > > > > On xextproto-7.0.2 the digest verification failed. Is there a change > > giong on or is there a bugy file on the server? > > > > The digest on the ebuild itself or a different file? If it's the > ebuild or something in the synced tree, try resyncing, and if that > doesn't work, > you can wait a day and try again, or verify against the file at > http://viewcvs.gentoo.org and redigest if you trust the results. > (Note that the viewcvs version won't exactly match either, or didn't > last I had > to use it, as its source tracking lines are slightly different. You > can verify the actual code, however, line by line or by downloading > and with a diff.) If the viewcvs version is the same but for the > source tracking lines, check for a bug and file one if there's none > filed. There's a known issue in instances when an ebuild was in the > tree (likely never unmasked), removed, and then later added again at > the same version, where > the system gets mixed up and the digest doesn't match. The size is > off by a specific small amount, 4 or 6 bytes, IIRC. That's the most > common reason for a no-match not attributable to a bad sync, and one > the Gentoo maintainer is often not aware of until he gets a bug about > it. > > If it's something in distfiles (basically, if it's one of the > tarballs), delete it from your distfiles cache and try again. It may > have been a problem in the download. If that doesn't fix it, check > bugs and file one > if necessary. > > FWIW, my last sync was a couple days ago (well, three, Sept. 5, early > morning US), but updated as of then, xextproto-7.0.2.ebuild has a > ctime of Feb 6, an mtime of Feb 4, so it has been around for awhile. > The Manifest file likewise, so no distfile changes since then, > either. I did > a total rebuild (emerge -e world) back in May (wow, has it been /that/ > long since gcc 4.2? seems so!), so that's when I last emerged it. > The tar.bz2 distfile should be 68323 bytes, the ebuild 444. > > Hmmm! "Houston. We have a problem!" > > I just synced to double-check, and while the version remained the same > and neither the ebuild nor the changelog changed, the Manifest did. > When > I looked at it above, it wasn't yet signed. It looks like they gpg- > signed it (a part of the security they are gradually implementing in > the tree), but when they did, something happened to the > distfile/tarball size. Above, it was 68323, now it says 68342, yet > the version number is the same! That should NOT happen! > > The previous one should I believe be the correct one. If you get > 68323 bytes and an md5sum of 242388ab65dde3a3dd313eeee265e429, > it /should/ be reasonably safe (but still it's your decision whether > the risk is worth it) to go ahead and redigest and merge it, as > that's probably the real one -- it agrees with what I have here. > > Looks like there's already a bug on it (from last year): > http://bugs.gentoo.org/show_bug.cgi?id=150225 > > Seems upstream (xorg) silently changed the tarball without changing > the version number... back in 2006. Maybe they pulled the same trick > once again (I see a passel of X updates waiting... on ~amd64, > probably not so many for stable... just checked, xorg 7.3 released on > the sixth, must be that). If so, it may be a bit before all sources > locations have the correct file, since the version didn't change, so > even deleting the tarball and redownloading might not get you the new > one for a few days. > > FWIW, deleting and redownloading, I get the 68323 byte version, same > as before. Maybe it's time for a new bug? Double-checking, yes, > it's time for a new bug, as downloading manually directly from (as > gotten from the ebuild, followed to the eclass): > > http://xorg.freedesktop.org/releases/individual/proto/ > > results in a file exactly 68323 bytes long, the old size. Thus, the > Manifest file seems to be wrong. > > OK, bug filed (with you credited as bringing it to my attention): > > http://bugs.gentoo.org/show_bug.cgi?id=191676 > Problem solved now, anyway. -- [EMAIL PROTECTED] mailing list
