view admin/notes/bugtracker @ 108488:925f002ec4ab

* iimage.el: Misc cleanup. (iimage-mode-map): Move initialization into declaration. (iimage-mode-buffer): Use with-silent-modifications. Simplify calling convention. Adjust callers. (iimage-mode): Don't run hook redundantly.

NOTES ON THE EMACS BUG TRACKER   -*- outline -*-

The Emacs Bug Tracker can be found at http://debbugs.gnu.org/

* Quick-start guide

This is 95% of all you will ever need to know.

** How do I report a bug?
Use M-x report-emacs-bug, or send mail to [e-mail].
If you want to Cc someone, use an "X-Debbugs-CC" header instead.

** How do I comment on a bug?
Reply to a mail on the bug-gnu-emacs list in the normal way.
Or send a mail to [e-mail].

If the bug is old and closed, you may have to unarchive it first.
Send a mail to [e-mail] with
unarchive 123
on the first line of the body.

** How do I close a bug?
Send a mail to [e-mail].  In the body, explain
why the bug is being closed.

** How do I set bug meta-data?
By mailing commands to [e-mail].  Place commands at the
start of the message body, one per line.

severity 123 serious|important|normal|minor|wishlist
tags 123 moreinfo|unreproducible|wontfix|patch

* More detailed information

For a list of all bugs, see http://debbugs.gnu.org/db/pa/lemacs.html
This is a static page, updated once a day.  There is also a dynamic
list, generated on request. This accepts various options, eg to see
the most recent bugs:

http://debbugs.gnu.org/cgi/pkgreport.cgi?newest=100

Or follow the links on the front page http://debbugs.gnu.org .

** How do I report a bug in Emacs now?
The same way as you always did.  Send mail to [e-mail],
or use M-x report-emacs-bug.

The only differences are:

i) Your report will be assigned a number and generate an automatic reply.

ii) Optionally, you can set some database parameters when you first
report a bug (see "Setting bug parameters" below).

iii) If you want to CC: someone, use X-Debbugs-CC: (this is important;
see below).

Once your report is filed and assigned a number, it is sent out to the
bug mailing list.  In some cases, it may be appropriate to just file a
bug, without sending out a copy.  To do this, send mail to
[e-mail].

** How do I reply to an existing bug report?
Reply to [e-mail], replacing 123 with the number
of the bug you are interested in.  NB this only sends mail to the
bug-list, it does NOT (?) send a CC to the original bug submitter.
So you need to explicitly CC him/her (and anyone else you like).

(Many people think the submitter SHOULD be automatically subscribed
to subsequent discussion, but this does not seem to be implemented.
See http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=37078)
See also http://debbugs.gnu.org/5439

Do NOT send a separate copy to the bug list address, since this may
generate a new report.  The only time to send mail to the bug list
address is to create a new report.

Gnus users can add the following to message-dont-reply-to-names;
similarly with Rmail and rmail-dont-reply-to-names:

"\\(emacs-pretest-bug\\|bug-gnu-emacs\\|bug-\\(e\\|gnu\\)macs\\)@gnu\\.org\\|\
\\(submit\\|control\\|owner\\)@debbugs\\.gnu\\.org"

The "[e-mail]" entry is there because it appears in the
"Resent-To" header.  For a long time Rmail erroneously included such
headers in replies.  If you correspond with an Rmail user on a bug,
these addresses may end up in the Cc.  Mailing to them does nothing
but create duplicates and errors.  (It is possible you might want to
have a dialog with the owner address, outside of normal bug
reporting.)

** When reporting a bug, to send a Cc to another address
(e.g. [e-mail]), do NOT just use a Cc: header.
Instead, use "X-Debbugs-CC:".  This ensures the Cc address will get a
mail with the bug report number in.  If you do not do this, each reply
in the subsequent discussion will end up creating a new bug.
This is annoying.

(So annoying that a form of message-id tracking has been implemented
to hopefully stop this happening, but it is still better to use X-Debbugs-CC.)

If a new report contains X-Debbugs-CC in the input, this is
converted to a real Cc header in the output.  (See Bug#1720).
It is also merged into the Resent-CC header (see below).

** How does Debbugs send out mails?

The mails are sent out to the bug list by being resent.  The From:
header is unchanged.  In new reports only (at present), the To:
address is altered as follows.  Any "bug-gnu-emacs",
"emacs-pretest-bug", or "submit@debbugs" address is replaced by
123@debbugs in the mail that gets sent out.  (This also applies to any
Cc: header, though you should be using X-Debbugs-CC instead in new
reports).  The original header is stored as X-Debbugs-Original-To, if
it was changed.  Any X-Debbugs-CC is merged into the Cc.

Mails arriving at the bug list have the following Resent-* headers:

Resent-From: person who submitted the bug
Resent-To:   [e-mail]
Resent-CC:   maintainer email address, plus any X-Debbugs-CC: entries

The "maintainer email address" is "[e-mail]" in most cases.

** To not get acknowledgement mail from the tracker,
add an "X-Debbugs-No-Ack:" header (with any value).  If you use Gnus,
you can add an element to gnus-posting-styles to do this automatically, eg:

("gnu-emacs\\(-pretest\\)?-bug"
   ("X-Debbugs-No-Ack" "yes"))

(adjust the regexp according to the name you use for the bug lists)

** To record a bug in the tracker without sending mail to the bug list.
This can be useful to make a note of something discussed on
emacs-devel that needs fixing.  In other words, this can be the
equivalent of adding something to FOR-RELEASE.

To: [e-mail]
[headers end]
Package: emacs
Version: 23.0.60
Severity: minor

Remember to fix FOO, as discussed on emacs-devel at http://... .

** Not interested in tracker control messages (tags being set, etc)?
Discard mails matching:

^X-GNU-PR-Message: (transcript|closed)

** How to avoid multiple copies of mails.
If you reply to reports in the normal way, this should work fine.
Basically, reply only to the numbered bug address (and any individual
people's addresses). Do not send mail direct to bug-gnu-emacs or
emacs-pretest-bug unless you are reporting a new bug.

** To close bug #123 (for example), send mail

To: [e-mail]

with a brief explanation in the body as to why the bug was closed.
There is no need to cc the address without the "-done" part or the
submitter; they get copies anyway so this will just result in more
duplicate mail.

** Details of closing a bug.
(For information only)
Sending a mail to 123-done does the following:

1) Mark the bug as closed in the database.

2) Send a mail to the original submitter telling them that their bug
has been closed.  This mail has a header:

X-GNU-PR-Message: they-closed 123

3) Send a mail to you and to the emacs-bug-tracker list confirming
that the bug has been closed.  This mail has a header:

X-GNU-PR-Message: closed 123

4) Send a copy of your mail to the bug-gnu-emacs list in exactly the
same way as if you had sent mail to "123" (sans -done). This mail has
headers:

X-GNU-PR-Message: cc-closed 123
Mail-Followup-To: [e-mail], person-who-closed

(This is Emacs-specific.  Normally the bug list gets the same mail as in 3).

** Setting bug parameters.
There are two ways to set the parameters of bugs in the database
(tags, severity level, etc).  When you report a new bug, you can
provide a "pseudo-header" at the start of the report, eg:

Package: emacs
Version: 23.0.60
Severity: minor

This can also include tags.  Some things (e.g. submitter) don't seem to
work here.

Otherwise, send mail to the control server, [e-mail].
At the start of the message body, supply the desired commands, one per
line:

command bug-number [arguments]
...
quit|stop|thank|thanks|thankyou|thank you

The control server ignores anything after the last line above.  So you
can place control commands at the beginning of a reply to a bug
report, and Bcc: the control server (note the commands have no effect
if you just send them to the bug-report number).  Bcc: is better than Cc:
in case people use Reply-to-All in response.

Some useful control commands:

*** To reopen a closed bug:
reopen 123

*** Bugs can be tagged in various ways (eg wontfix, patch, etc).
The available tags are:
patch wontfix moreinfo unreproducible fixed notabug
See http://debbugs.gnu.org/Developer#tags
The list of tags can be prefixed with +, - or =, meaning to add (the
default), remove, or reset the tags. E.g.:

tags 123 + wontfix

** URL shortcuts

http://debbugs.gnu.org/...

123             # given bug number
123;mbox=yes    # mbox version of given bug
package         # bugs in given package
from:[e-mail]
severity:severity      # all bugs of given severity
tag:tag                # all bugs with given tag

** Usertags

See <http://wiki.debian.org/bugs.debian.org/usertags>

"Usertags" are very similar to tags: a set of labels that can be added
to a bug.  There are two differences between normal tags and user tags:

1) Anyone can define any valid usertag they like.  In contrast, only a
limited, predefined set of normal tags are available (see above).

2) A usertag is associated with a specific email address.

You set usertags in the same way as tags, by talking to the control
server.  One difference is that you can also specify the associated
email address.  If you don't explicitly specify an address, then it
will use the one from which you send the control message. The address
must have the form of an email address (with an "@" sign and least 4
characters after the "@").

*** Setting usertags

a) In a control message:

user [e-mail]
usertags 1234 any-tag-you-like

This will add a usertag "any-tag-you-like" to bug 1234.  The tag will
be associated with the address "[e-mail]".  If you omit
the first line, the tag will be associated with your email address.

The syntax of the usertags command is the same as that of tags (eg wrt
the optional [=+-] argument).

b) In an initial submission, in the pseudo-header:

User: [e-mail]
Usertags: a-new-tag

Again, the "User" is optional.

*** Searching by usertags

The search interface is not as advanced as for normal tags.  You need
to construct the relevant url yourself rather than just typing in a
search box.  The only piece you really need to add is the "users"
portion, the rest has the same syntax as normal.

**** To browse bugs by usertag:
http://debbugs.gnu.org/cgi/pkgindex.cgi?indexon=users

**** To find all bugs usertagged by a given email address:

http://debbugs.gnu.org/cgi/pkgreport.cgi?users=[e-mail]

(Supposedly, the "users" field can be a comma-separated list of more
than one email address, but it does not seem to work for me.)

**** To find bugs tagged with a specific usertag:

This works just like a normal tags search, but with the addition of a
"users" field.  Eg:

http://debbugs.gnu.org/cgi/pkgreport.cgi?users=[e-mail];tag=calendar

*** To merge bugs:
Eg when bad replies create a bunch of new bugs for the same report.
Bugs must all be in the same state (e.g. same package(s) and severity
-- see `reassign' and `severity' below), but need not have the same
tags (tags are merged). E.g.:

merge 123 124 125 ...

Note that merging does not affect titles.  In particular, a "retitle"
of merged bugs only affects individual bugs, not all of them.

*** Forcing a merge:
Like `merge', but bugs need not be in the same state.  The packages
must still match though (see `reassign' below).  The first one listed
is the master.  E.g.:

forcemerge 123 124 125 ...

Note: you cannot merge with an archived bug - you must unarchive it first.

*** To unmerge bugs:
To disconnect a bug from all bugs it is merged with:

unmerge 123

This command accepts only one bug number.

*** To clone bugs:
Useful when one report refers to more than one bug.

clone 123 -1 [-2 ...]
retitle -1 second bug
retitle -2 third bug

The negative numbers provide a way to refer to the cloned bugs (which
will be assigned proper numbers).

NB you cannot clone a merged bug.  You'd think that trying to do so
would just give you an unmerged copy of the specified bug number, but no:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=474742

You must unmerge, clone, then re-merge.

*** To set severity:
severity 123 critical|grave|serious|important|normal|minor|wishlist

See http://debbugs.gnu.org/Developer#severities for the meanings.

*** To set the owner of a bug:
owner 123 A Hacker <[e-mail]>

The shorthand `!' means your own address.

*** To remove the owner of a bug:
noowner 123

*** To mark a bug as fixed in a particular version:
fixed 123 23.0.60

*** To remove a "fixed" mark:
notfixed 123 23.0.60

*** To assign or reassign a bug to a package or list of packages:
reassign 1234 emacs

** To remove spam from the tracker, move it to the `spam' pseudo-package:
reassign 123 spam

** To change the title of a bug:
retitle 123 Some New Title

** To change the submitter address:
submitter 123 [e-mail]

Note that it does not seem to work to specify "Submitter:" in the
pseudo-header when first reporting a bug.

** How does archiving work?
You can still send mail to a bug after it is closed.  After 28 days with
no activity, the bug is archived, at which point no more changes can
be made.  If you try to send mail to the bug after that (or merge with
it), it will be rejected.  To make any changes, you must unarchive it first:

unarchive 123

The bug will be re-archived after the next 28 day period of no activity.

** The web-page with the list of bugs is slow to load

It's a function of the number of displayed bugs.  You can speed things
up by only looking at the newest 100 bugs:
http://debbugs.gnu.org/cgi-bin/pkgreport.cgi?newest=100;package=emacs

Or use the static index:
http://debbugs.gnu.org/db/ix/full.html

** What are those "mbox folder" links on the bug report pages?

"mbox folder" = messages as they arrived at the tracker

"status mbox" = as above, but with a fake message at the start
    summarizing the bug status

"maintainer mbox" = messages as sent out from the tracker to the
    maintainers (ie, bug-gnu-emacs).  These have some changed headers
    (Resent-*, Subject, etc).

** What do the pkgreport.cgi sort options mean?

"normal" = by open/closed status, then severity, then tag, then bug number

"oldview" = as above, but without the tag part

"age" = as normal, but sort in decreasing order of last modification
time, rather than by increasing bug number

"raw" = ?

** ChangeLog issues

*** When you fix a bug, it can be helpful to put the bug number in the
ChangeLog entry, for example:

   * foo.el (foofunc): Fix the `foo' case.  (Bug#123)

Then the relevant bug can be found for easy reference.  If it's an
obvious fix (e.g. a typo), there's no need to clutter the log with the
bug number.

Similarly, when you close a bug, it can be helpful to include the
relevant ChangeLog entry in the message to the bug tracker, so people
can see exactly what the fix was.

*** bug-reference-mode

Activate `bug-reference-mode' in ChangeLogs to get clickable links to
the bug web-pages.

*** Debian stuff

http://lists.gnu.org/archive/html/emacs-devel/2009-11/msg00440.html

** Bazaar stuff

*** You can use `bzr commit --fixes emacs:123' to mark that a commit fixes
Emacs bug 123.  You will first need to add a line to your bazaar.conf:

bugtracker_emacs_url = http://debbugs.gnu.org/{id}

Note that all this does is add some metadata to the commit, it doesn't
actually mark the bug as closed in the tracker.  There seems to be no
way to see this "metadata" with `bzr log', which is rather poor, but
it will show up as a link in a recent loggerhead installation, or with
some of the graphical frontends to bzr log.

** Gnus-specific voodoo

*** Put point on a bug-number and try: M-x gnus-read-ephemeral-emacs-bug-group

*** If the above is not available:
(add-hook 'gnus-article-mode-hook
          (lambda ()
             (setq bug-reference-url-format "http://debbugs.gnu.org/%s")
              (bug-reference-mode 1)))

and you can click on the bug number in the subject header.


* Technical Notes

The following are technical notes on how it works.  These are just for
reference, you don't need to read these as a user of the system.

Getting mail from the Emacs bug list into the tracker requires the
assistance of sysadmin at gnu.org.  The test tracker set-up was, I
think, [gnu.org #359140]:
http://lists.gnu.org/archive/html/savannah-hackers/2008-03/msg00074.html
http://lists.gnu.org/archive/html/savannah-hackers/2008-04/msg00034.html

** The debbugs.gnu.org setup was handled in [gnu.org #510605].
There are two pieces (replace AT with @ in the following):

i) fencepost has an /etc/aliases entry:
emacs-pretest-bug: submit AT debbugs.gnu.org

ii) An exim router:
emacsbugs_router:
  driver = redirect
  senders = !Debian-debbugs AT debbugs.gnu.org
  local_parts = bug-gnu-emacs
  domains = gnu.org
  data = submit AT debbugs.gnu.org

This says, for mail arriving at bug-gnu-emacs, only allow it through
to the list if it was sent from debbugs.gnu.org.  Otherwise, send
it to the submit address at the bug-tracker.

FIXME There's probably an issue with the mail-news gateway here that
still needs to be addressed (bug#936).

** fencepost's /etc/exim4/local_domains configuration needs a line
!debbugs.gnu.org adding [gnu.org #503532].  Otherwise people on
fencepost can't report bugs, since *.gnu.org addresses are assumed to
be handled locally on fencepost, unless otherwise specified.

** All mail arriving at debbugs.gnu.org is first run through SpamAssassin.
Obvious spam is rejected, the rest is sent on to the moderated list
debbugs-submit.  Approved mail is passed on to the tracker.
(Note this means that messages may appear out of sequence in the
tracker, since mail from whitelisted senders goes straight through.)

NOTE: An alternative to this would be to use listhelper AT nongnu.org
as a moderator address.  Eg the emacs-bug-tracker list uses this.
It does basic spam processing on the moderator requests and
automatically rejects the obviously bogus ones.  Someone still has to
accept the good ones though.  The advantage of this would not be having
to run and tune our own spam filter.  See
http://savannah.nongnu.org/projects/listhelper

An "X-Debbugs-Envelope-To" header is used to keep track of where the
mail was actually bound for:
http://lists.gnu.org/archive/html/emacs-devel/2009-11/msg01211.html

** Mailing list recipient/sender filters.
The following mailman filters are useful to stop messages being
needlessly held for moderation:

*** debbugs-submit
(quiet|control|submit)@(debbugs\.gnu\.org|emacsbugs\.donarmstrong\.com)
[0-9]+(-done|-quiet|-subscribe)?@(debbugs\.gnu\.org|emacsbugs\.donarmstrong\.com)
(bug-gnu-emacs|emacs-pretest-bug|bug-(e|gnu)macs)@gnu\.org

bug-emacs and bug-gnumacs are lesser-used aliases from fencepost's
/etc/aliases file.

*** emacs-bug-tracker
sender: bug-gnu-emacs AT gnu.org
recipient: emacs-bug-tracker AT debbugs\.gnu\.org

The latter is because that is the address that debbugs actually sends to.
An /etc/aliases entry redirects it to the real emacs-bug-tracker address.

** Recovering from moderation mistakes

All discarded messages are stored in /var/lib/mailman/spam.
If a non-spam message accidentally gets discarded, just do:

cat /var/lib/mailman/spam/not-really-spam.msg | /usr/lib/debbugs/receive
... check it works ...
mv /var/lib/mailman/spam/not-really-spam.msg /var/lib/mailman/not-spam/

** Administrivia

The debbugs-submit list should have the administrivia option off,
else it can by mistake filter out requests to subscribe to bugs.
But, this feature doesn't work anyway (see bug#5439).

** How to test changes

Add an entry to /etc/debbugs/Maintainers like:

mytest       my.email.address

Then if you do all your testing with 'Package: mytest', the resulting
mails should only go to your email address.