- 06:38 Ticket #123 (scripts-remove works poorly with non-Athena accounts) created by
- It's unclear that we should care, but scripts-remove handles being given …
- 06:24 Ticket #122 (scripts-remove doesn't delete Rails databases) created by
- If you autoinstall a Rails app and then remove it, you'll still be left …
- 06:20 Ticket #121 (Uninstalling applications can destroy un-associated user data) created by
- If I have a database with user data that is unassociated with an …
- 06:14 Ticket #120 (autoinstalled scripts should be uninstallable without interacting with ...) created by
- At the moment, if I don't know my sql.mit.edu password, I can't uninstall …
- 06:06 Ticket #86 (Running multiple autoinstalled instances of Trac causes them all to look ...) closed by
- fixed: This seems to no longer be an issue. Recent Trac autoinstalls seem to use …
- 05:48 Ticket #119 (scripts-remove shouldn't sign users up for sql.mit.edu) created by
- As far as I can tell, scripts-remove will sign users up for sql.mit.edu if …
- 05:01 Ticket #118 (Autoinstallers should be more controllable from the command-line) created by
- It would be neat if you could pre-answer a bunch of the autoinstaller …
- 04:28 Changeset  by
- People with DB's want to know their uninstall worked, too
- 04:26 Changeset  by
- Remove /~$lname/ from a couple more places
- 04:14 Changeset  by
- "What's this /~$lname/ thing?" Provide a prompt that talks about http://$lname.scripts.mit.edu/ instead of http://scripts.mit.edu/~$lname/, since the latter has been dead for about a year.
- 00:12 Changeset  by
- Put the initial user crontab under revision control
- 22:44 Ticket #85 (Trac autoinstaller should configure trac.ini properly when using git) closed by
- fixed: Fixed by Alex in r1512, merged to trunk in r1514 and deployed.
- 22:39 Changeset  by
- Fix Trac's git browser plugin in autoinstalls Closes Trac #85 (Merge of r1511:1512 from branches/locker-dev to trunk, from adehnert)
- 22:31 Ticket #113 (Django auto-installs on a machine where USER != ATHENA_USER get wrong ...) closed by
- fixed: Merged to trunk in r1513 and deployed to the locker.
- 22:26 Changeset  by
- Make autoinstallers use ATHENA_USER instead of USER Closes Trac #113 (Merge of r1510:1511 from branches/locker-dev to trunk, from adehnert)
- 03:40 Ticket #95 (Onserver scripts hard-code deploy directory) closed by
- fixed: Yup.
- 20:29 Ticket #3 (Interface for automatic vhost management) closed by
- fixed: Yeah, I think we can close this now.
- 20:22 Changeset  by
- Fix Trac's git browser plugin in autoinstalls
- 19:10 Changeset  by
- Make autoinstallers use ATHENA_USER instead of USER Closes Trac #113
- 01:18 Changeset  by
- It's a new thing! It's a nice thing! It's... Supe^WOpenAFS 1.4.12!
- 23:51 Ticket #117 (Review and deploy Haskell static-cat) created by
- A Haskell version static-cat has been …
- 04:13 Changeset  by
- Return real error codes from unauthorized PGetTokens, PSetTokens, PUnlog. The previous return values of 0 caused aklog to segfault.
- 02:13 Changeset  by
- Don't remove groups on uninstallation of nss_nonlocal
- 01:18 Changeset  by
- Temporarily scriptsify shadow-utils to fix max length of group names The upstream shadow-utils package used to hardcode the max length of a group name to 16 characters, and Fedora patched it to make that length match utmpx.ut_name's length (32) and documented it as being 32. The last upstream update that Fedora took changed the code to default to 16, but allow a nondefault value via a --with flag to configure. Fedora's patch thus no longer applied and they dropped the relevant hunk, but neglected to add the --with flag in the spec file, so the max group name length dropped back to 16, causing packages that install groups with names longer than 16 characters to begin failing to install properly mid-release. We'll be pursuing this issue with Fedora.
- 23:23 Changeset  by
- New CellServDB from grand.central.org, via Ops
- 06:29 Changeset  by
- CVE-2010-0434: Putting the "Patch" in "Apache" since... well, 2010
- 03:24 Changeset  by
- Fix graph titles...
- 00:05 Changeset  by
- New scripts-wizard package This package presently provides a symlink for wizard in /usr/local/bin so that users have it in their paths all the time.
- 22:58 Changeset  by
- Directory for scripts-wizard package Yes, it's empty. Yes, it's expected to stay that way. No, I can't build the package before it's committed. Yes, this is silly.
- 13:39 Changeset  by
- We won't always have paris (CellServDB update, take 2)
- 13:10 Changeset  by
- Updated CellServDB from Ops
- 06:09 Changeset  by
- Upgrade Apache from 2.2.13-1 to 2.2.14-1
- 05:22 Changeset  by
- Make sure that SRPMs for upstream packages actually come from upstream We don't usually put SRPMs for packages we build in our yum repo since we generate the SRPM from a (possibly patched) version of the upstream spec and source files contained in the upstream SRPM as part of our build process. However, occasionally an SRPM or two of ours find their way into our repo. If we pull in that SRPM, a subsequent build of that package will fail because most of our patch hunks will fail due to having already been applied. Avoid this screw case by never downloading the allegedly upstream SRPMs from our own repo.
- 02:25 Changeset  by
- Support out-of-locker autoinstallers This parameterizes the locker that the autoinstaller lives in, as well as the contact info for the autoinstaller's maintainer. (Merge of r1491:1496 from branches/locker-dev to trunk, originally from broder)
- 01:18 Changeset  by
- Don't let people shoot themselves in the foot by leaving ailocker set to ""
- 01:18 Changeset  by
- Don't attach the scripts locker twice.
- 01:15 Changeset  by
- Distinguish between scripts for scripts and scripts for autoinstallers. Some pieces of the autoinstaller process are clearly tied to scripts itself - e.g. scripts to get information from sql.mit.edu. Others are clearly tied to the autoinstaller - e.g. the /mit/scripts/deploy/bin/foo scripts. Out-of-tree autoinstallers want to replace the latter, but not the former. For scripts that fall into the former category, explicitly call them from /mit/scripts. For those that fall into the latter, call them from /mit/$ailocker. New out-of-tree autoinstallers can define ailocker to provide their own hooks.
- 01:01 Changeset  by
- Use a slightly more concise shell expression for setting variables.
- 00:14 Changeset  by
- Parametrize the maintainer and contact for autoinstallers. This allows the Scripts autoinstaller infrastructure to be leveraged by out-of-tree autoinstallers.
- 18:33 Changeset  by
- Make it possible for Accounts to create users on scripts (new moira, again) (Yes, you misread that.)
- 03:10 Changeset  by
- New moira packages
- 23:44 Changeset  by
- Fix directions for adding static extensions ========= Instance: 1149427 Time: Mon Feb 8 20:29:45 2010 Host: BYTE-ME.MIT.EDU From: Busy inodes after unmount. Self-destruct in 5 seconds. Have a nice day. [FAILED] <mitchb> Also, those directions need to be taken out back and shot. We don't do things like that anymore. ========= Done.
- 23:21 Changeset  by
- Various tweaks to rails autoreloading code o Don't watch all of public, as apps may cache files there and cause the fcgi to repeatedly die and destroy any fcgi performance gains o Specifically watch .htaccess and dispatch.fcgi o Raise an error to die instead of trying to reload code and dying in the process because you can't o Explain that killing the fcgi process results in a reload of the code at the next request (Merge of r1411:1486 from branches/locker-dev to trunk, originally from gdb)
- 23:00 Changeset  by
- Renewed certificate for schuh.mit.edu
- 03:47 Changeset  by
- Explain why we raise an error here
- 03:37 Changeset  by
- Time to be honest with ourselves. We're not reloading anything here.
- 03:31 Changeset  by
- Watch specific files as well
- 03:09 Changeset  by
- __scripts/needcerts: Add support for working around Safari Safari on Mac OS X (or more properly, CFNetwork and the rest of the SSL stack) doesn't properly support SSLVerifyClient Optional, which is our default for :444. In particular, if you don't have an identity preference set, only SSLVerifyClient Require will trigger the dialog to set an identity preference and present a certificate to the site: http://lists.apple.com/archives/apple-cdsa/2009/Apr/msg00041.html We can work around this by checking for the Safari user-agent in /__scripts/needcerts and renegotiating SSLVerifyclient Require. Forcing the Require behavior on Safari users that reach this page is reasonable because this page is only (supported to be) reached as an ErrorDocument 401; if you're intentionally using AuthOptional on to take advantage of the optional authentication, you'll never trigger the 401 error.
Note: See TracTimeline for information about the timeline view.