Age | Commit message (Collapse) | Author |
|
We were not properly running dedupe on cached threads, so fix it by only
caching post-dedupe messages.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
When "b4 am" or "b4 mbox" is passed an arbitrary URL that matches a
public-inbox scheme but that doesn't match the midmask URL, assume the
user knows what they are doing and attempt to retrieve the mbox.
Suggested-by: Philippe Blain <levraiphilippeblain@gmail.com>
Link: https://msgid.link/f2b0f3c1-d98d-979d-1513-18f2491f4267@gmail.com
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
When we're adding a mismatched trailer, tell us why it hasn't been
accepted.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Looks like at some point we started ignoring the -p flag, so restore
this to the expected functionality
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
It is now possible to create a b4-tracked branch from an arbitrary
thread (or from a previously sent b4-tracked series):
b4 prep -F [msgid-of-the-series]
Example:
$ b4 prep -F 20220901194310.115427-1-tony.luck@intel.com
Grabbing thread from lore.kernel.org/all/20220901194310.115427-1-tony.luck%40intel.com/t.mbox.gz
Checking attestation on all messages, may take a moment...
---
✓ [PATCH 1/3] EDAC/skx_common: Use driver decoder first
✓ [PATCH 2/3] EDAC/skx_common: Make output format similar
✓ [PATCH 3/3] EDAC/i10nm: Add driver decoder for Ice Lake and Tremont CPUs
---
✓ Signed: DKIM/intel.com
---
Created new branch b4/edac_improve_memory
Applying 3 patches
---
Applying: EDAC/skx_common: Use driver decoder first
Applying: EDAC/skx_common: Make output format similar
Applying: EDAC/i10nm: Add driver decoder for Ice Lake and Tremont CPUs
---
NOTE: any follow-up trailers were ignored; apply them with b4 trailers -u
This makes it easier to start tracking pre-existing series with b4 prep.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
If a link contains a '@', we wrongly consider it a person-trailer.
Ignore any URL-containing trailers that we don't explicitly recognize.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Fixes regression introduced by trailers refactor.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Add a missing continue that caused incorrectly reported trailer
mismatches.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
With the addition of b4 trailers it became pretty obvious that the way
we originally implemented trailers didn't age well. This refactor does
the following:
- introduces LoreTrailer class to replace passing trailers as tuples
- reimplements trailer-order with strict adherence to chain-of-custody
rules
- adds tests to most common trailer follow-up/ordering cases
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
The git-send-email configuration allows smtpserver to be defined as a
path to a local sendmail-like command. Implement the same logic in b4,
allowing fuller compatibility with git-send-email.
Suggested-by: Maxime Ripard <maxime@cerno.tech>
Link: https://msgid.link/20220830122217.h52zswusnlb3iggq@meerkat.local
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
The sendemail sections are supposed to be fall-throughs where a named
section overrides the values provided by the global section. This
implements the required logic to be compliant with git.
Suggested-by: Rob Herring <robh@kernel.org>
Link: https://msgid.link/20220825182506.1449442-1-robh@kernel.org
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
We could be sending the patches via a web endpoint, for which we don't
need to force CRLF line endings. Convert into CRLF only immediately
before sending the message out via smtp.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
For UX reasons, make --resend a separate switch instead of operating on
the --prefixes RESEND logic.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Force the use of smtp server when sending anything other than patches
(e.g. b4 ty automated responses).
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
This implements most of the web endpoint submission functionality.
Completely undocumented and needs a lot more testing before it's useful,
but we're getting close.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
We're not going to care about untracked files for most of b4 prep
operations, so don't error out when they are present and we're running
"is the tree clean" checks.
Suggested-by: Maxime Ripard <maxime@cerno.tech>
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
The scripts expect to be running from the toplevel directory, so make
sure to switch back to git topdir before running get_maintainer.pl
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
First go at implementing the tip-commit strategy. It shares a lot with
the 'commit' strategy, but there are gotchas for situations where the
cover letter commit is suddenly not the tip commit any more (rebase, new
commits, etc).
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Another, hopefully final overhaul of commands and flags:
- "b4 ez-series" is now "b4 prep"
- "b4 ez-trailers" is now "b4 trailers"
- "b4 ez-send" is now "b4 send"
I've also split on-disk output into two different commands:
b4 prep --format-patch <outdir>: does not set To/Cc and doesn't do any From
magic. In effect, it's as close as it gets to git format-patch output
compatibility.
b4 send --dry-run -o <outdir>: generates the messages exactly as they
are about to be sent, then writes them out to the directory specified.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Fix a logic bug that prevented the To: header from being added on
--dry-run and do a better job making sure the addresses are valid and
sane.
Reported-by: Mattijs Korpershoek <mkorpershoek@baylibre.com>
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
I need to refactor the entire approach to trailers so we stop passing
around tuples. It made sense at the time, but now it's just a source of
bugs.
Reported-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Significant refactor of (formerly) "b4 submit" based on initial
feedback:
1. Split "b4 submit" into three different commands:
- ez-series: for managing the series cover letters, tracking info, etc
- ez-trailers: for retrieving trailers and updating commits (works on
any branch, not just ez-series branches)
- ez-send: for sending branches managed by ez-series
2. Refactor to support multiple cover letter strategies:
- the default "commit" strategy that keeps the cover letter in an
empty commit (should be backwards-compatible with "b4 submit")
- the non-invasive "branch-description" strategy that keeps the cover
letter in the branch.branchname.description configuration setting and
tracking in branch.branchname.b4-tracking
- the not-yet-implemented "tag" strategy that mimics the behaviour of
git-publish
The strategy can be set via b4.ez-cover-strategy variable, e.g. in
your .gitconfig:
[b4]
ez-cover-strategy = branch-description
Note, that converting from one strategy to another doesn't work and
will probably explode in weird ways right now.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
git-rev-list returns patches in reverse chronological order by default,
which is the opposite of what we need. Add a --reverse to the call to
get the expected behaviour.
Reported-by: Mattijs Korpershoek <mkorpershoek@baylibre.com>
Link: https://lore.kernel.org/tools/87ilnti947.fsf@baylibre.com/
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
When using set_payload(), we should pass along the charset as well,
otherwise we run into trouble when converting to/from bytes again.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
This is the first rough implementation of "b4 submit". Currently
implemented:
- b4 submit --new : to start a new branch
- b4 submit --edit-cover : to edit the cover message
- b4 submit --update-trailers : to receive latest trailer updates from
the mailing lists
- b4 submit --send : sends the messages using existing git.sendemail
configs
For details, see "b4 submit --help".
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
The trailer *names* should always be ascii. While it is possible to
imagine that some project would have non-ascii trailer names like:
Signalé-par: Developpeur Un <d1@example.com>
Co-développé-avec: Developpeur Deux <d2@example.com>
Nonetheless, I am not aware of any project doing this, and this allows
us to weed out a bunch of false-positives from bad forwards.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Chain-of-custody order requires that Signed-off-by closes the list of
added trailers.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Per discussion on the mailing lists, reordering trailers is almost never
the right decision, so remove support for trailer ordering completely.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
We don't want to use the set() here, since we want to preserve the
order, so use an auxiliary set for dupe tracking.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
A lot of maintainers use patchwork alongside b4, to make it easier to
track patches and rely on some CI integration. This commit adds some
basic patchwork integration:
- on "b4 am", "b4 shazam", "b4 pr" we will mark the relevant patchwork
entries as "Under Review"
- on "b4 ty" we can set these patches as "Accepted"
- on "b4 ty -d" we can set them as "Deferred"
To make it work, the following entries must be present in the repository
used with b4:
[b4]
pw-key = (your API token)
pw-url = https://patchwork.kernel.org
pw-project = (your project, e.g. linux-usb)
pw-review-state = under-review
pw-accept-state = accepted
pw-discard-state = deferred
To get your patchwork API token, go to your patchwork profile page.
The pw-accept-state and pw-discard-state can be overridden using the
--pw-set-state flag to "b4 ty". E.g. if you wanted to mark the patches
as "Not applicable":
b4 ty -d 5 --pw-set-state not-applicable
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
It is time to roll out 0.9.0 for wider use. Update requirements to the
latest supported and tested versions.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
It is a common request to be able to get a partial thread in case
someone submitted an auxiliary standalone patch in the middle of a
larger patch series. Passing the msgid of the start of the thread along
with --no-parent should tell b4 to break the thread at the start of the
message-id specified and only consider that message and its children.
Suggested-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/tools/YpTI9lhCfA7shi6j@sirena.org.uk/
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Minor nitpicking by PyCharm, but not completely unreasonable.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Sometimes the DKIM record will have the full email address instead of
just the @domainname.com part. For those cases, drop the local part so
we don't falsely claim domain mismatching during b4-am runs.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
When performing attestor identity comparisons, lowercase email addresses
and domain names for case-insensitive matching.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
b4's usage of git-log '--branches' option is broken. The option takes a
glob pattern *only* and must have an '=', but b4 ends up passing
'--branches <guessbranch>' to git-log. This will kind of work, but is
not checking only 'guessbranch'. For example, these 3 commands all do
something different:
git log -1 --branches=master
git log -1 --branches master
git log -1 --branches=*aster
A maintainer wanting to apply a patch or series likely has a small set of
known branches they apply patches to. Using a glob pattern is not a good
fit for that. Instead, allow --guess-branch to be repeated and to take
fixed refs.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
Link: https://lore.kernel.org/r/20220331195346.1384515-1-robh@kernel.org
|
|
On lore.kernel.org we provide a unified index of all mailing lists in
/all/, which removes the need to perform a redirect lookup when querying
by message-id. However, some public-inbox instances may not have that,
so we still need to be able to fall back to that redirect lookup. Adapt
a patch from Rob Herring to support both situations.
Link: https://lore.kernel.org/tools/20220225031135.4136158-1-robh@kernel.org/
Suggested-by: Rob Herring <robh@kernel.org>
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Sometimes folks indent their trailers with whitespace, so relax the
rules to recognize this situation. In theory, this shouldn't introduce
false-positives.
Suggested-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20211214214327.4003631-1-robh@kernel.org
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Instead of hardcoding the logic to bail at all presence of control
characters, soften it a bit to only error out when we see the presence
of Cfs in the absence of any other character from a non-latin character
set in the same line.
Additionally, show exactly where the Cfs were found when printing out
the error message.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Implement initial support for checking if the patch message contains
unicode control characters that can be used to trick code reviewer into
accepting maliciously formatted code.
Link: https://lore.kernel.org/tools/20211101175020.5r4cwmy4qppi7dis@meerkat.local/
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
There appears to be a bug in smtplib that doesn't properly support 8-bit
content even when upstream SMTP gateway supports 8BITMIME (they all do,
it's not 90s any more). Work around this by passing the message payload
as bytes instead of string.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
I'm felling comfortable that "b4 ty" is sufficiently mature at this
point to implement sending thank-yous directly. This is only the initial
implementation that covers only the very basic parts of git's sendemail
configuration options, but this should actually cover 90% of cases if
not more.
One important caveat -- I moved the "b4 ty -s" flag to be "b4 ty -t" in
order to disambiguate it from the capital -S (that actually does the
sending). Since "b4 ty" is still marked as an experimental feature, I
feel we can do this without much impact.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
If b4 doesn't have an explicit keyringsrc configured, attempt to fetch
it from patatt configuration instead. This avoids some very confusing
situations where patatt would validate and b4 would reject. (I had an
expired pubkey in my gpg default keyring, but the same pubkey had an
updated expiration date in patatt's keyring.)
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
Link: https://lore.kernel.org/r/20211007165551.2312946-1-keescook@chromium.org
|
|
Allow people to set up their own preferred merge templates, using the
netdev standard as default.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
By popular demand, provide a way to apply series straight to a git
repository. By default, we're still going the safest possible route:
- create a sparse worktree consisting just of the files being modified
- run "git am" against the temporary worktree
- if "git am" went well, fetch from the temporary worktree into our
current tree and leave everything in FETCH_HEAD
- unless we're running "b4 shazam -A" in which case we just apply to the
current HEAD (exact equivalent of b4 am -o- | git am)
Further changes to come based on feedback.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
New version of lore.kernel.org is live today, so release the version of
b4 that works best with it.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
When generating am-ready patch series, separate each standard body part
with a single pair of newlines regardless of how many the original
message contained.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Extindex allows us to get rid of a couple of kludges:
- we no longer need to manually backfill, as /all/ contains all sources
- we can just query /all/ for new series
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
We want to dedupe all threads we retrieve from public-inbox, so do this
in the central place instead of only when doing get_strict_tread().
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Instead of relying on DKIM validation, use list-id preference when
dealing with multiple messages matching the same message-id. We may
end up adding an attestation check to it as well in the future.
Signed-off-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|