Gene,
Neither Dustin nor me had the privilege to fix the permission issue.
Permission get fixed, I uploaded the latest snapshot.
Jean-Louis
Gene Heskett wrote:
On Sunday 28 February 2010, Dustin J. Mitchell wrote:
On Sat, Feb 27, 2010 at 7:34 PM, Gene Heskett <gene.hesk...@verizon.net>
wrote:
I am sorry Dustin, but the permissions problem alluded to earlier does
not seem to be fixed yet so the old link still shows the 20100131
versions as the newest, and the 'download tarball' buttons on all the
sourceforge pages at the link above, while bringing up a requester asking
what should firefox do with this file, with the save button already
checked, do nothing when clicking on the ok. No file is downloaded.
If I understand, you're talking about the "Download GNU Tarball" link
on the SourceForge ViewVC pages. The link works for me, but even so
it's not a very effective way to download snapshots: the resulting
tarballs will not be the usual "distribution" tarballs, as they have
not had their ./autogen scripts run. If you'd like, I can help you
modify your test scripts to build straight from subversion instead of
>from snapshots.
Breakage that continues for over 3 weeks now, does lead to questions, and
the replies seem to intend to placate, but have done nothing of substance
to restore our ability to continue our near daily testing of the bleeding
edge.
I don't mean to placate - we screwed up. But to be fair, you only
brought it up at 9pm on a Friday! I'm sure it'll be sorted out soon.
I used to see more updates on the weekends, and figured you both had other,
buys the bread, duties during the week.
I don't know exactly how Jean-Louis makes the snapshot tarballs, but
I've made a distribution tarball built from the latest subversion,
which should be similar, available here:
http://djmitche.s3.amazonaws.com/amanda-2.6.2alpha.tar.gz
Let me know if you have any trouble making it work.
Just one niggle. It was not named internally in the tarball with the
snapshot date. My build & install script depends on that, so now I have a
2.6.2alpha install that when it comes time to do a cleanup of old versions,
something I do about monthly, it will need a careful check of its file dates
else I might remove the wrong library or some such. My script also cleans
out any srcs it finds, and without the snapshot date, doesn't leave an old
version behind so I have a ready made, backup to the previous tree left in
/home/amanda that to recover to it, is a simple cd into the older directory
followed by a make install as root. I tried to rename the tarball but that
resulted in the tar.gz's deletion so I had to go get it 3 times. It did
install, and amcheck ran normally, so we'll see how tonight's run goes.
Needless to say, if I do this again, I'll overwrite this one, which isn't a
'Good Thing(TM)' so I will now rename that tree with today's date to prevent
that.
And that of course makes me wonder if I am indeed the only person doing
any test builds and actual use of that test build at all. I certainly
hope not.
I hope not, too! Certainly the more people who test Amanda, the more
quickly and effectively we will catch and solve bugs. I'm a big
believer in testing, both automated and manual. It *is* disconcerting
that nobody (not even you?) noticed this for, as you say, three weeks.
With luck, this thread will encourage some new folks to begin
regularly testing Amanda, especially the upcoming release.
"Given enough eyeballs, all bugs are shallow" [1]
Dustin
[1] http://en.wikipedia.org/wiki/Linus'_Law
My test scripts are attached. Quite simple. Not even a kilobyte combined.
Syntax is:
./newmanada amandatarball-snapshotdate (without the .tar.gz) (as root)
