5 - Symlinked rerere to get awesomeness. Problems:
7 - Might not make a huge difference; how does it handle empty file
8 and removed file cases?
9 - Need to manually run git rerere subsequently to reap benefits
10 - Majority of resolutions have to happen pre-merge (see below)
11 - Consider workflow: run wizard mass-upgrade, and then begin
12 resolving working copies one by one. Each time we resolve
13 a copy, it should cause other copies to start magically resolving.
14 So, ordering should be:
16 2. If it fails, merge the rr-cache with central rr-cache
17 (this operation needs to be atomic) and replace it
18 with a symlink. File permissions preferably should
19 be made correct, but don't have to be since only root
20 will be touching subsequently. If the hash already exists,
21 don't do anything (maybe record this for the benefit
22 of Mister Kite aka so we don't have to do a full traversal,
23 this optimization might be essential)
24 3. When a human is resolving the merges, they are "low
25 concurrency", i.e. only one commit recording rerere will
26 happen at a time. This means that rr-cache does not
27 need to be concurrent safe. Some number of hashes in
28 the rr-cache will start having postimages; we'll use
29 a full-scan to figure that out. Then cross-reference those
30 with the recorded pending resolutions, and figure out which
31 checkouts we can run rerere on (this gets permissions kind
32 of tricky). We'll try an alternative plan: manually require
33 the user run some sort of retry command that does this as
34 root; presumably they'd run this every ten installs or
35 something. A user can run git rerere to get a resolution
37 This requires some new data-structures:
38 - Besides the merge.txt file (which should never ever change),
39 we should have an outstanding.txt file which gets modified
40 as our scripts do resolutions behind our back. Those modifications
41 might a little annoying for a human to keep up with, so we should
42 recommend something like watch -n2 "head file" or something
43 - We need to keep track of the hashes and the cross-referencing.
44 A very small sqlite database might be a good idea here, although
45 the type of information we're interested in a somewhat unnatural
46 query. Alternatively, we just have a very simple text file.
47 - Make it possible to say certain classes of missing files are ok
49 - Wizard needs a correct arch/ setup
50 - The wizard command, when not on scripts, should automatically SSH to
51 scripts and start executing there?
52 - Write the code to make Wordpress figure out its URL from the database
54 - Remerges aren't reflected in the parent files, so `git diff` output is
55 spurious. Not sure how to fix this w/o tree hackery.
56 - Sometimes users remove files. Well, if those files change, they automatically
57 get marked as conflicted. Maybe we should say for certain files "if they're
58 gone, they're gone forever"? What is the proper resolution?
60 - Parse output HTML for class="error" and give those errors back to the user (done),
61 then boot them back into configure so they can enter in something different
63 - Replace gaierror with a more descriptive name (this is a DNS error)
65 - Pre-emptively check if daemon/scripts-security-upd
66 is not on scripts-security-upd list (/mit/moira/bin/blanche)
67 - If you try to do an install on scripts w/o sql, it will sign you up but fail to write
68 the sql.cnf file. This sucks.
70 - Web application for installing autoinstalls has a hard problem
71 with credentials (as well as installations that are not conducted
72 on an Athena machine.) We have some crazy ideas involving a signed
73 Java applet that uses jsch to SSH into athena.dialup and perform
77 - Tidy up common code in callAsUser and drop_priviledges in shell;
78 namely cooking up the sudo and environment variable lines
79 - Summary script should be more machine friendly, and should not
80 output summary charts when I increase specificity
81 - Report code in wizard/command/__init__.py is ugly as sin. Also,
82 the Report object should operate at a higher level of abstraction
83 so we don't have to manually increment fails. (in fact, that should
84 probably be called something different). The by-percent errors should
86 - Move resolutions in mediawiki.py to a text file? (the parsing overhead
88 - PHP end of file allows omitted semicolon, can result in parse error
89 if merge resolutions aren't careful. `php -l` can be a quick stopgap
92 - Figure out why Sphinx sometimes fails to crossref :func: but wil
93 crossref :meth:, even though the dest is very clearly a function.
94 Example: :func:`wizard.app.php.re_var`
95 - The TODO extension for Sphinx doesn't properly force a full-rebuild
97 - Make single user mass-migrate work when not logged in as root. The
98 primary difficulty is making the parallel-find information easily
99 accessible to individual users: perhaps we can do a single-user
100 parallel-find on the fly.
101 - Don't use the scripts heuristics unless we're on scripts with the
102 AFS patch. Check with `fs sysname`
103 - Make 'wizard summary' generate nice pretty graphs of installs by date
104 (more histograms, will need to check actual .scripts-version files.)
105 - It should be able to handle installs like Django where there's a component
106 that gets installed in web_scripts and another directory that gets installed
109 - ACLs is a starting point for sending mail to users, but it has
110 several failure modes:
111 - Old maintainers who don't care who are still on the ACL
112 - Private AFS groups that aren't mailing lists and that we
114 A question is whether or not sending mail actually helps us:
115 many users will probably have to come back to us for help; many
116 other users won't care.
118 PULLING OUT CONFIGURATION FILES IN AN AUTOMATED MANNER
120 advancedpoll: Template file to fill out
121 django: Noodles of template files
122 gallery2: Multistage install process
123 joomla: Template file
124 mediawiki: One-step install process
125 phpbb: Multistage install process
126 phpical: Template file
129 wordpress: Multistage install process
131 COMMIT MESSAGE FIELDS:
133 Installed-by: username@hostname
134 Pre-commit-by: Real Name <username@mit.edu>
135 Upgraded-by: Real Name <username@mit.edu>
136 Migrated-by: Real Name <username@mit.edu>
137 Wizard-revision: abcdef1234567890
138 Wizard-args: /wizard/bin/wizard foo bar baz
142 Committer: Real Name <username@mit.edu>
143 Author: lockername locker <lockername@scripts.mit.edu>
147 - It is not required nor expected for update scripts to exist for all
148 intervening versions that were present pre-migration; only for it
149 to work on the most recent migration.
151 - Currently all repositories are initialized with --shared, which
152 means they have basically ~no space footprint. However, it
153 also means that /mit/scripts/wizard/srv MUST NOT lose revs after
158 * Some parts of the infrastructure will not be touched, although I plan
159 on documenting them. Specifically, we will be keeping:
161 - parallel-find.pl, and the resulting
162 /mit/scripts/.htaccess/scripts/sec-tools/store/scriptslist
164 * The new procedure for generating an update is as follows:
165 (check out the mass-migration instructions for something in this spirit,
166 although uglier in some ways; A indicates the step /should/ be automated)
168 0. ssh into not-backward, temporarily give the daemon.scripts-security-upd
169 bits by blanching it on system:scripts-security-upd, and run parallel-find.pl
171 1. [ see doc/upgrade.rst ]
173 [ENTER HERE FROM CREATING A NEW REPO]
175 9. Push all of your changes in a public place, and encourage others
176 to test, using --srv-path and a full path.
178 [ XXX: doc/deploy.rst ]
179 GET APPROVAL BEFORE PROCEEDING ANY FURTHER;
180 THIS IS PUSHING THE CHANGES TO THE PUBLIC
182 NOTE: The following commands are to be run on not-backward.mit.edu.
183 You'll need to add daemon.scripts-security-upd to
184 scripts-security-upd to get bits to do this. Make sure you remove
185 these bits when you're done.
187 10. Run `wizard research appname`
188 which uses Git commands to check how many
189 working copies apply the change cleanly, and writes out a logfile
190 with the working copies that don't apply cleanly. It also tells
191 us about "corrupt" working copies, i.e. working copies that
192 have over a certain threshold of changes.
194 11. Run `wizard mass-upgrade appname`, which applies the update to all working
197 12. Run parallel-find.pl to update our inventory
199 [ XXX: doc/upgrade.rst ]
200 * For mass importing into the repository, there are a few extra things:
202 * When mass producing updates, if the patch has changed you will have to
203 do a special procedure for your merge:
205 git checkout pristine
206 # NOTE: Now, the tricky part (this is different from a real update)
207 git symbolic-ref HEAD refs/heads/master
208 # NOTE: Now, we think we're on the master branch, but we have
209 # pristine copy checked out
210 # NOTE: -p0 might need to be twiddled
211 patch -p0 < ../app-1.2.3/app-1.2.3.patch
213 # reconstitute .scripts directory
214 git checkout v1.2.2-scripts -- .scripts
216 # NOTE: Fake the merge
217 git rev-parse pristine > .git/MERGE_HEAD
219 You could also just try your luck with a manual merge using the patch
222 [ XXX: doc/layout.rst ]
223 * The repository for a given application will contain the following files:
225 - The actual application's files, as from the official tarball
227 - A .scripts directory, with the intent of holding Scripts specific files
228 if they become necessary.
230 - .scripts/dsn, overriding database source name