summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorGravatar Joey Hess <joey@kitenet.net>2014-04-23 17:10:05 -0400
committerGravatar Joey Hess <joey@kitenet.net>2014-04-23 17:10:05 -0400
commit1883edd4e9fd0fff00494d1c39a984095c3a45d1 (patch)
tree28bdacc668074a78563d557232296446369d0057
parent7cc2dd1c86573d555a5ee57ede362ec871a919e9 (diff)
devblog
-rw-r--r--doc/devblog/day_157__upgrade_checking.mdwn22
1 files changed, 22 insertions, 0 deletions
diff --git a/doc/devblog/day_157__upgrade_checking.mdwn b/doc/devblog/day_157__upgrade_checking.mdwn
new file mode 100644
index 000000000..ff70b8093
--- /dev/null
+++ b/doc/devblog/day_157__upgrade_checking.mdwn
@@ -0,0 +1,22 @@
+Now git-annex's self-upgrade code will check the gpg signature of a
+new version before using it.
+
+To do this I had to include the gpg public keys into the
+git-annex distribution, and that raised the question of which public keys
+to include. Currently I have both the dedicated git-annex distribution
+signing key, and my own gpg key as a backup in case I somehow misplace the
+former.
+
+Also spent a while looking at the recent logs on the web server. There
+seem to be around 600 users of the assistant with
+upgrade checking enabled. That breaks down to 68% Linux amd64, 20% Linux
+i386, 11% OSX Mavericks, and 0.5% OSX Lion.
+
+Most are upgrading successfully, but there are a few that seem to
+repeatedly fail for some reason. (Not counting the OSX Lion, which will
+probably never find an upgrade available.) I hope that someone who is
+experiencing an upgrade failure gets in touch with some debug logs.
+
+In the same time period, around 450 unique hosts manually downloaded a
+git-anex distribution. Also compare with Debian popcon, which has 1200
+reporting git-annex users.