]> Sergey Matveev's repositories - public-inbox.git/log
public-inbox.git
5 years agoTODO: support migration to v2 without breaking v1 "git fetch"
Eric Wong [Mon, 4 Feb 2019 22:37:32 +0000 (22:37 +0000)]
TODO: support migration to v2 without breaking v1 "git fetch"

I'd like to move https://public-inbox.org/git/ to v2; but
cronjobs using "git fetch" to following should not break.

5 years agoviewdiff: group path match to not be confused by "/dev/null"
Eric Wong [Mon, 4 Feb 2019 11:40:13 +0000 (11:40 +0000)]
viewdiff: group path match to not be confused by "/dev/null"

Leaving out parentheses caused transitions to state="del" or
state="add" to be misidentified.

cf. https://public-inbox.org/meta/20190204105454.GG10587@szeder.dev/

Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
5 years agonewswww: add /$MESSAGE_ID global redirector endpoint
Eric Wong [Wed, 9 Jan 2019 11:43:27 +0000 (11:43 +0000)]
newswww: add /$MESSAGE_ID global redirector endpoint

This is the fallback for the normal WWW endpoint.

Adding this to the top-level seems to be alright, since lynx and
w3m both understand nntp://<HOSTNAME>/<Message-ID> anyways.

If newsgroup and inbox names conflict, then consider it the
fault of the original sender.

Since NewsWWW is intended to support buggy linkifiers in mail clients,
they can interpret nntp:// URLs as http://<HOSTNAME>/<Message-ID>

Inbox ordering from the config file is preserved since
commit cfa8ff7c256e20f3240aed5f98d155c019788e3b
("config: each_inbox iteration preserves config order"),
so admins can rely on that to configure how scanning
works.

Requested-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
cf. https://public-inbox.org/meta/20190107190719.GE9442@pure.paranoia.local/
    nntp://news.public-inbox.org/20190107190719.GE9442@pure.paranoia.local

5 years agoviewdiff: support renames and long paths in diffstat anchors
Eric Wong [Fri, 1 Feb 2019 07:36:02 +0000 (07:36 +0000)]
viewdiff: support renames and long paths in diffstat anchors

This is best-effort, but works well-enough in practice for
projects which use shell-friendly filenames as well as the
long path names for some Linux kernel selftests.

5 years agoviewdiff: escape HTML ampersand for renames
Eric Wong [Fri, 1 Feb 2019 06:30:17 +0000 (06:30 +0000)]
viewdiff: escape HTML ampersand for renames

For URLs we generate, we need to escape '&' in query parameters
for correctness.

5 years agoviewvcs: allow '0' as a valid filename for blob downloads
Eric Wong [Fri, 1 Feb 2019 06:06:04 +0000 (06:06 +0000)]
viewvcs: allow '0' as a valid filename for blob downloads

Only to be pedantic...

5 years agolinkify: support proto://hostname without trailing slash
Eric Wong [Fri, 1 Feb 2019 05:27:52 +0000 (05:27 +0000)]
linkify: support proto://hostname without trailing slash

Sometimes users will write "http://example.com" without the
trailing slash, which every browser and tool I've tested seems
to understand.

5 years agoview: simplify quote splitting
Eric Wong [Fri, 1 Feb 2019 03:06:47 +0000 (03:06 +0000)]
view: simplify quote splitting

Perl "split" can capture and group in the regexp itself,
so rely on that to shorten our code.

Comparing the /T/ HTML output of a thread from hell (on LKML with
1356 messages) reveals no difference in the rendered result.
Only the HTML source differs in newline placement before/after
the closing </span>

This allows a minor speedup on my X32 Thinkpad @ 1.6GHz with
the aforementioned LKML thread from hell:

before: 3.67s
 after: 3.55s

5 years agoview: fix broken hunk header hrefs in Atom feeds
Eric Wong [Fri, 1 Feb 2019 01:47:31 +0000 (01:47 +0000)]
view: fix broken hunk header hrefs in Atom feeds

We use absolute URLs in the Atom feeds (to ease
syndication/mirroring), so hunk headers need to point to the
solver URLs.

5 years agoview: diffstat anchors for multi-message/attachment views
Eric Wong [Thu, 31 Jan 2019 11:33:49 +0000 (11:33 +0000)]
view: diffstat anchors for multi-message/attachment views

diffstat <-> ^diff anchors work within the same attachment or
message while in HTML views which display multiple messages.

5 years agoviewdiff: diffstat links to diff anchors
Eric Wong [Thu, 31 Jan 2019 10:45:55 +0000 (10:45 +0000)]
viewdiff: diffstat links to diff anchors

This can be helpful for reviewing larger patches which span
across several files on the permalink (/$MESSAGE_ID/) HTML
page.

More work will be needed to get this working for the /T/ and /t/
pages which show multiple emails, as the filename-based anchors
will conflict at the moment.

5 years agohval: routines for attribute escaping
Eric Wong [Mon, 18 Jan 2016 21:16:14 +0000 (21:16 +0000)]
hval: routines for attribute escaping

We'll use HTML attributes + anchor links to link to filenames
in coming commits.

5 years agoMerge remote-tracking branch 'origin/purge'
Eric Wong [Thu, 31 Jan 2019 21:08:48 +0000 (21:08 +0000)]
Merge remote-tracking branch 'origin/purge'

* origin/purge:
  implement public-inbox-purge tool
  v2writable: read epoch on purge
  v2writable: cleanup processes when done
  v2writable: purge ignores non-existent git epoch directories
  v2writable: ->purge returns undef on no-op
  import: purge: reap fast-export process
  hoist out resolve_repo_dir from -index

5 years agodoc: remove completed TODO items
Eric Wong [Thu, 31 Jan 2019 09:04:12 +0000 (09:04 +0000)]
doc: remove completed TODO items

5 years agodoc/config: document "replyto" configuration knob
Eric Wong [Thu, 31 Jan 2019 09:17:45 +0000 (09:17 +0000)]
doc/config: document "replyto" configuration knob

I hate it, but it's necessary to support some mirrors.

5 years agodoc/config: user documentation for limiters
Eric Wong [Thu, 31 Jan 2019 08:52:49 +0000 (08:52 +0000)]
doc/config: user documentation for limiters

I've relied on this feature to keep the VPS behind
https://public-inbox.org/git/ from OOM-ing since 2016,
so document it to ensure others can make use of low-end
servers like I do.

More limiters may become configurable for viewvcs and
solver functionality (or we continue using the default
one).

5 years agoconfig: tiny cleanup to use _array() sub
Eric Wong [Thu, 31 Jan 2019 08:28:41 +0000 (08:28 +0000)]
config: tiny cleanup to use _array() sub

5 years agoqspawn: documentation updates
Eric Wong [Thu, 31 Jan 2019 08:17:15 +0000 (08:17 +0000)]
qspawn: documentation updates

This will become critical for future changes to display
git commits, diffs, and trees.

Use "qspawn.wcb" instead of "qspawn.response" to enhance
readability.

5 years agoinbox: drop psgi.url_scheme requirement from base_url
Eric Wong [Thu, 31 Jan 2019 06:00:22 +0000 (06:00 +0000)]
inbox: drop psgi.url_scheme requirement from base_url

This will make it easier to make command-line tools
from SolverGit.

5 years agoviewvcs: support streaming large blobs
Eric Wong [Thu, 31 Jan 2019 04:10:29 +0000 (04:10 +0000)]
viewvcs: support streaming large blobs

Forking off git-cat-file here for streaming large blobs is
reasonably efficient, at least no worse than using
git-http-backend for serving clones.  So let our limiter
framework deal with it.

git itself isn't great for large files, and AFAIK there's no
stable/widely-available mechanisms for reading smaller chunks
of giant blobs in git itself.

Tested with some giant GPU headers in the Linux kernel.

5 years agosolvergit: allow shorter-than-necessary OIDs from user
Eric Wong [Thu, 31 Jan 2019 02:46:13 +0000 (02:46 +0000)]
solvergit: allow shorter-than-necessary OIDs from user

We can rely on git to disambiguate, here; because sometimes
shorter OIDs can be unambiguous even if we only resolved the
longer one.

5 years agosolvergit: allow searching on longer-than-needed OIDs
Eric Wong [Thu, 31 Jan 2019 01:41:15 +0000 (01:41 +0000)]
solvergit: allow searching on longer-than-needed OIDs

public-inbox can only index the abbreviated object_ids in
emails, not the full or even longer-than-necessary object_ids.
So retry failed object_ids if they're longer than 7 hex
characters.

5 years agoinbox: perform cleanup of Git objects for coderepos
Eric Wong [Thu, 31 Jan 2019 00:32:42 +0000 (00:32 +0000)]
inbox: perform cleanup of Git objects for coderepos

Otherwise, long-running but idle git processes may keep unlinked
packs around indefinitely and waste disk space.

5 years agot/config.t: test PublicInbox::Git sharing between inboxes
Eric Wong [Wed, 30 Jan 2019 20:48:53 +0000 (20:48 +0000)]
t/config.t: test PublicInbox::Git sharing between inboxes

We need to ensure we don't introduce unnecessary processes
and memory usage for mapping multiple inboxes to the same
code repos.

5 years agodoc/config: document coderepo and css bits
Eric Wong [Wed, 30 Jan 2019 11:38:52 +0000 (11:38 +0000)]
doc/config: document coderepo and css bits

New features ought to be documented

5 years agosolvergit: don't confuse Xapian with ".." in filenames
Eric Wong [Wed, 30 Jan 2019 10:33:55 +0000 (10:33 +0000)]
solvergit: don't confuse Xapian with ".." in filenames

Xapian will interpret ".." as ranges, even quoted phrases.
So break up words on ".." since punctuation (AFAIK) is not
searchable, anyways.

5 years agogit: use "git rev-parse --git-path"
Eric Wong [Wed, 30 Jan 2019 08:47:43 +0000 (08:47 +0000)]
git: use "git rev-parse --git-path"

Using git worktrees was causing t/solver_git.t to fail on me.

5 years agoMerge remote-tracking branch 'origin/viewvcs' into master
Eric Wong [Wed, 30 Jan 2019 07:40:35 +0000 (07:40 +0000)]
Merge remote-tracking branch 'origin/viewvcs' into master

* origin/viewvcs: (66 commits)
  solvergit: deal with alternative diff prefixes
  solvergit: extract mode from diff headers properly
  solvergit: avoid "Wide character" warnings
  solvergit: do not show full path names to "git apply"
  css/216dark: add comments and tweak highlight colors
  viewvcs: avoid segfault with highlight.pm at shutdown
  solvergit: do not solve blobs twice
  t/check-www-inbox: disable history
  t/check-www-inbox: don't follow mboxes
  t/check-www-inbox: replace IPC::Run with PublicInbox::Spawn
  hval: add src_escape for highlight post-processing
  viewvcs: wire up syntax-highlighting for blobs
  hlmod: disable enclosing <pre> tag
  t/hl_mod: extra check to ensure we escape HTML
  wwwhighlight: read_in_full returns undef on errors
  solver: crank up max patches to 9999
  viewvcs: do not show final error message twice
  qspawn: decode $? for user-friendliness
  solver: reduce "git apply" invocations
  solver: hold patches in temporary directory
  ...

5 years agoview: remove unused _msg_date sub
Eric Wong [Wed, 30 Jan 2019 07:38:51 +0000 (07:38 +0000)]
view: remove unused _msg_date sub

Not needed since commit 956abe9ad5f13a0d1755262be412d6a54fda72e9
("view: depend on SearchMsg for Message-ID")

5 years agohttpd: a few comments about some fields we set
Eric Wong [Wed, 30 Jan 2019 07:39:02 +0000 (07:39 +0000)]
httpd: a few comments about some fields we set

Removing 'psgix.input.buffered' could be a possibility in
the future.

5 years agosolvergit: deal with alternative diff prefixes
Eric Wong [Wed, 30 Jan 2019 04:38:26 +0000 (04:38 +0000)]
solvergit: deal with alternative diff prefixes

At least, without extra directory levels, since
git-diff supports --src-prefix and --dst-prefix,
and /git/6aa8857a11/s/ uses it...

5 years agosolvergit: extract mode from diff headers properly
Eric Wong [Wed, 30 Jan 2019 04:29:24 +0000 (04:29 +0000)]
solvergit: extract mode from diff headers properly

grep() won't set $1, so use "=~", instead.

5 years agosolvergit: avoid "Wide character" warnings
Eric Wong [Wed, 30 Jan 2019 03:31:12 +0000 (03:31 +0000)]
solvergit: avoid "Wide character" warnings

Just quiet Perl down, since we don't know or care about the
encoding of the patch we hand off to git-apply.

5 years agosolvergit: do not show full path names to "git apply"
Eric Wong [Wed, 30 Jan 2019 02:32:29 +0000 (02:32 +0000)]
solvergit: do not show full path names to "git apply"

"git apply" will warn about whitespace with the full path of the
patch, which will expose the $TMPDIR environment to users over
HTTP(S).

This change breaks compatibility with git pre-1.8.5, again;
but that was released in late-2013; so hopefully everybody
is on newer versions.

5 years agocss/216dark: add comments and tweak highlight colors
Eric Wong [Tue, 29 Jan 2019 22:31:26 +0000 (22:31 +0000)]
css/216dark: add comments and tweak highlight colors

Overkill, but "highlight" supports single-line comments (slc)
independently of multi-line comments (com); but we'll use the
same color for that.

We'll also use #0f0 instead of #0ff for "kwb" (keyword class "b")
since blue shades are prevalent in <a> links and comments, while
green was unused.

5 years agoviewvcs: avoid segfault with highlight.pm at shutdown
Eric Wong [Tue, 29 Jan 2019 21:35:25 +0000 (21:35 +0000)]
viewvcs: avoid segfault with highlight.pm at shutdown

Proper ordering of destruction seems required to avoid segfaults
at shutdown.

5 years agosolvergit: do not solve blobs twice
Eric Wong [Tue, 29 Jan 2019 04:01:21 +0000 (04:01 +0000)]
solvergit: do not solve blobs twice

In some cases, a file may ping-pong between blob IDs in the same
message when reverts occur.  So break out of this early.

This doesn't account for different abbreviations, but the
limited variations of abbreviations should alleviate the
problem.

5 years agot/check-www-inbox: disable history
Eric Wong [Tue, 29 Jan 2019 03:42:48 +0000 (03:42 +0000)]
t/check-www-inbox: disable history

WWW::Mechanize keeps an infinitely large stack, which was
leading to OOM errors on my system.

5 years agot/check-www-inbox: don't follow mboxes
Eric Wong [Tue, 29 Jan 2019 03:37:24 +0000 (03:37 +0000)]
t/check-www-inbox: don't follow mboxes

They can be extremely large with no limit, so can lead to OOM
errors.

5 years agot/check-www-inbox: replace IPC::Run with PublicInbox::Spawn
Eric Wong [Mon, 28 Jan 2019 09:02:42 +0000 (09:02 +0000)]
t/check-www-inbox: replace IPC::Run with PublicInbox::Spawn

Because WWW::Mechanize uses truckload of memory, fork
needs to prepare all that memory for CoW, which ends up
bailing with ENOMEM.

5 years agomid: filter out 'y', 'n', and email addresses from references()
Eric Wong [Tue, 29 Jan 2019 07:44:28 +0000 (07:44 +0000)]
mid: filter out 'y', 'n', and email addresses from references()

Looking at git@vger history, several emails had broken
References/In-Reply-To pointing to <y>, <n> and email
addresses as Message-IDs in References and In-Reply-To
headers.

This was causing too many unrelated messages to be linked
together in the same thread.

5 years agohval: add src_escape for highlight post-processing
Eric Wong [Sun, 27 Jan 2019 11:43:56 +0000 (11:43 +0000)]
hval: add src_escape for highlight post-processing

We need to post-process "highlight" output to ensure it doesn't
contain odd bytes which cause "wide character" warnings or
require odd glyphs in source form.

5 years agoviewvcs: wire up syntax-highlighting for blobs
Eric Wong [Sun, 27 Jan 2019 11:38:42 +0000 (11:38 +0000)]
viewvcs: wire up syntax-highlighting for blobs

And update 216dark.css to match a color scheme I'm used to;
which is fairly minimal and doesn't use all the classes
"highlight" provides.

5 years agohlmod: disable enclosing <pre> tag
Eric Wong [Sun, 27 Jan 2019 11:36:22 +0000 (11:36 +0000)]
hlmod: disable enclosing <pre> tag

We already have a <pre> tag in ViewVCS, and nesting <pre>
inside the pre-existing <pre> overrides the "white-space:pre"
we use to align line numbers.

5 years agot/hl_mod: extra check to ensure we escape HTML
Eric Wong [Sun, 27 Jan 2019 11:35:29 +0000 (11:35 +0000)]
t/hl_mod: extra check to ensure we escape HTML

Otherwise, it's open season on our users :<

5 years agowwwhighlight: read_in_full returns undef on errors
Eric Wong [Sun, 27 Jan 2019 05:09:22 +0000 (05:09 +0000)]
wwwhighlight: read_in_full returns undef on errors

The return value of "print" is not undef for Perl IO::Handle.

5 years agosolver: crank up max patches to 9999
Eric Wong [Sun, 27 Jan 2019 03:38:41 +0000 (03:38 +0000)]
solver: crank up max patches to 9999

Might as well, since the only constraint is filesystem space
for temporary files for public-inbox-httpd users.

-httpd can fairly share work across clients with our use of
psgi_qx; and there's a recent patch series in git@vger with 64
patches in sequence.

5 years agoviewvcs: do not show final error message twice
Eric Wong [Sun, 27 Jan 2019 03:32:01 +0000 (03:32 +0000)]
viewvcs: do not show final error message twice

SolverGit::ERR already writes the exception to the debug
log before calling {user_cb}, so there's no need for viewvcs
to append it.

5 years agoqspawn: decode $? for user-friendliness
Eric Wong [Sun, 27 Jan 2019 03:30:30 +0000 (03:30 +0000)]
qspawn: decode $? for user-friendliness

The raw value of $? isn't very useful, generally.

5 years agosolver: reduce "git apply" invocations
Eric Wong [Sun, 27 Jan 2019 00:34:35 +0000 (00:34 +0000)]
solver: reduce "git apply" invocations

"git apply" is capable of applying multiple patches in one
invocation, so give it multiple patches on the command-line
now that we no longer rely on anonymous file handles to hold
patches.

This cuts down a 64-patch series on git@vger from ~1s to ~800ms
with vfork spawn enabled using Inline::C.

5 years agosolver: hold patches in temporary directory
Eric Wong [Sun, 27 Jan 2019 00:21:51 +0000 (00:21 +0000)]
solver: hold patches in temporary directory

We can avoid bumping up RLIMIT_NOFILE too much by storing
patches in a temporary directory.  And we can share this
top-level directory with our temporary git repository.

Since we no longer rely on a working-tree for git, we are free
to rearrange the layout and avoid relying on the ".git"
convention and relying on "git -C" for chdir.

This may also ease porting public-inbox to older systems
where git does not support "-C" for chdir.

5 years agosolver: rewrite to use Qspawn->psgi_qx and pi-httpd.async
Eric Wong [Tue, 22 Jan 2019 02:10:13 +0000 (02:10 +0000)]
solver: rewrite to use Qspawn->psgi_qx and pi-httpd.async

The psgi_qx routine in the now-abandoned "repobrowse" branch
allows us to break down blob-solving at each process execution
point.  It reuses the Qspawn facility for git-http-backend(1),
allowing us to limit parallel subprocesses independently of Perl
worker count.

This is actually a 2-3% slower a fully-synchronous execution;
but it is fair to other clients as it won't monopolize the server
for hundreds of milliseconds (or even seconds) at a time.

5 years agoview: swap CRLF for LF in HTML output
Eric Wong [Sat, 26 Jan 2019 20:05:56 +0000 (20:05 +0000)]
view: swap CRLF for LF in HTML output

It makes no difference to browsers aside from saving a few
bytes; and this means we won't have to worry about extra
'%0D' showing up in links to solver.

5 years agot/qspawn.t: psgi_qx stderr test
Eric Wong [Tue, 22 Jan 2019 07:57:06 +0000 (07:57 +0000)]
t/qspawn.t: psgi_qx stderr test

5 years agoqspawn: implement psgi_qx
Eric Wong [Wed, 18 Jan 2017 07:27:03 +0000 (07:27 +0000)]
qspawn: implement psgi_qx

This new asynchronous API, will allow us to take
advantage of non-blocking I/O from even small commands;
as those may still need to wait for slow operations.

5 years agohttpd/async: stop running command if client disconnects
Eric Wong [Fri, 13 Jan 2017 23:10:25 +0000 (23:10 +0000)]
httpd/async: stop running command if client disconnects

If an HTTP client disconnects while we're piping the output of a
process to them, break the pipe of the process to reclaim
resources as soon as possible.

5 years agoqspawn|httpd/async: improve and fix out-of-date comments
Eric Wong [Wed, 11 Jan 2017 04:12:27 +0000 (04:12 +0000)]
qspawn|httpd/async: improve and fix out-of-date comments

5 years agoqspawn|getlinebody: support streaming filters
Eric Wong [Wed, 11 Jan 2017 04:12:26 +0000 (04:12 +0000)]
qspawn|getlinebody: support streaming filters

This is intended for wrapping "git show" and "git diff"
processes in the future and to prevent it from monopolizing
callers.

This will us to better handle backpressure from gigantic
commits.

5 years agoqspawn: implement psgi_return and use it for githttpbackend
Eric Wong [Sun, 25 Dec 2016 08:53:19 +0000 (08:53 +0000)]
qspawn: implement psgi_return and use it for githttpbackend

Was: ("repobrowse: port patch generation over to qspawn")

We'll be using it for githttpbackend and maybe other things.

5 years agohttpd/async: remove needless sysread wrapper
Eric Wong [Sun, 8 Jan 2017 04:31:30 +0000 (04:31 +0000)]
httpd/async: remove needless sysread wrapper

We don't appear to be using it anywhere

5 years agot/check-www-inbox: trap SIGINT for File::Temp destruction
Eric Wong [Mon, 21 Jan 2019 06:51:23 +0000 (06:51 +0000)]
t/check-www-inbox: trap SIGINT for File::Temp destruction

Otherwise, temporary GDBM files don't get unlinked
when I SIGINT the process.

5 years agohval: split out escape sequences to a separate table
Eric Wong [Mon, 21 Jan 2019 03:30:56 +0000 (03:30 +0000)]
hval: split out escape sequences to a separate table

We'll want to handle those escape sequences independently,
"highlight" already does HTML escaping.

5 years agohighlight: initial wrapper and PSGI service
Eric Wong [Mon, 21 Jan 2019 03:19:20 +0000 (03:19 +0000)]
highlight: initial wrapper and PSGI service

I'll probably expose the PSGI service for cgit;
but it could be useful to others as well.

5 years agot/check-www-inbox: warn on missing Content-Type
Eric Wong [Mon, 21 Jan 2019 03:09:22 +0000 (03:09 +0000)]
t/check-www-inbox: warn on missing Content-Type

Oops, I might've left it out, somewhere.

5 years agoconfig: each_inbox iteration preserves config order
Eric Wong [Sun, 20 Jan 2019 20:56:22 +0000 (20:56 +0000)]
config: each_inbox iteration preserves config order

For cross-inbox Message-ID resolution; having some sort of
stable ordering makes the most sense.  Relying on the
order of the config file seems most natural and allows us
to avoid introducing yet another configuration knob.

5 years agosolver: remove extra "^index $OID..$OID" line
Eric Wong [Sun, 20 Jan 2019 13:48:10 +0000 (13:48 +0000)]
solver: remove extra "^index $OID..$OID" line

It was harmless, besides wasting space and memory.

5 years agosolver: force quoted-printable bodies to LF
Eric Wong [Sun, 20 Jan 2019 13:46:28 +0000 (13:46 +0000)]
solver: force quoted-printable bodies to LF

..if the Email::MIME ->crlf is LF.

Email::MIME::Encodings forces everything to CRLF on
quoted-printable messages for RFC-compliance; and
git-apply --ignore-whitespace seems to miss a context
line which is just "\r\n" (w/o leading space).

5 years agoviewvcs: disable white-space prewrap in blob view
Eric Wong [Sun, 20 Jan 2019 12:18:35 +0000 (12:18 +0000)]
viewvcs: disable white-space prewrap in blob view

We need to keep line-numbers from <a> tags synced to the actual
line numbers in the code when working in smaller viewports.

Maybe I only work on reasonable projects, but excessively
long lines seem to be less of a problem in code than they are
in emails.

5 years agoviewdiff: do not link to 0{7,40} blobs (again)
Eric Wong [Sun, 20 Jan 2019 12:16:47 +0000 (12:16 +0000)]
viewdiff: do not link to 0{7,40} blobs (again)

We must reset diff context when starting a new file;
and we must check for all-zeroes object_ids as the
post-image correctly.

5 years agoextmsg: don't bother partial matching with <16 chars
Eric Wong [Sun, 20 Jan 2019 12:33:20 +0000 (12:33 +0000)]
extmsg: don't bother partial matching with <16 chars

It's not worth it, and attempts to wildcard off
single-character Message-IDs(*) causes Xapian to error
out in unpredictable ways:

  something terrible happened at /usr/lib/x86_64-linux-gnu/perl5/5.24/Search/Xapian/Enquire.pm line 54.
    ...propagated at lib/PublicInbox/Search.pm line 209.

So don't bother.

(*) because people blindly hit 'y' or 'n' when git-send-email
    prompted them for In-Reply-To.

5 years agot/check-www-inbox: use xmlstarlet to validate Atom if available
Eric Wong [Sun, 20 Jan 2019 11:40:22 +0000 (11:40 +0000)]
t/check-www-inbox: use xmlstarlet to validate Atom if available

I almost forgot about this script; but remembering to test
it against real-world data can be useful to hunt for bugs.

5 years agoviewdiff: quote attributes for Atom feed
Eric Wong [Sun, 20 Jan 2019 10:57:13 +0000 (10:57 +0000)]
viewdiff: quote attributes for Atom feed

We still need to use XHTML the Atom feed, and XHTML requires
attributes to be quoted, whereas HTML 5 does not.

5 years agoviewdiff: cleanup state transitions a bit
Eric Wong [Sun, 20 Jan 2019 10:42:07 +0000 (10:42 +0000)]
viewdiff: cleanup state transitions a bit

This makes things less error-prone and allows us to only
highlight the "@@ -\S+ \+\S+ @@" part of the hunk header
line, without highlighting the function context.

This more closely matches the coloring behavior of git-diff(1)

5 years agoviewdiff: support diff-highlighting w/o coderepo
Eric Wong [Sun, 20 Jan 2019 09:42:34 +0000 (09:42 +0000)]
viewdiff: support diff-highlighting w/o coderepo

Having diff highlighting alone is still useful, even
if blob-resolution/recreation is too expensive or
unfeasible.

5 years ago$INBOX/_/text/color/ and sample user-side CSS
Eric Wong [Sun, 20 Jan 2019 09:29:08 +0000 (09:29 +0000)]
$INBOX/_/text/color/ and sample user-side CSS

Since we now support more CSS classes for coloring,
give this feature more visibility.

5 years agowww: admin-configurable CSS via "publicinbox.css"
Eric Wong [Sun, 20 Jan 2019 04:21:07 +0000 (04:21 +0000)]
www: admin-configurable CSS via "publicinbox.css"

Maybe we'll default to a dark theme to promote energy savings...

See contrib/css/README for details

5 years agosolver: restore diagnostics and deal with CRLF
Eric Wong [Sat, 19 Jan 2019 21:57:26 +0000 (21:57 +0000)]
solver: restore diagnostics and deal with CRLF

Apparently Email::MIME returns quoted-printable text
with CRLF.  So use --ignore-whitespace with git-apply(1)
and ensure we don't capture '\r' in pathnames from
those emails.

And restore "$@" dumping when we die while solving.

5 years agoview: enforce trailing slash for /$INBOX/$OID/s/ endpoints
Eric Wong [Sat, 19 Jan 2019 08:54:42 +0000 (08:54 +0000)]
view: enforce trailing slash for /$INBOX/$OID/s/ endpoints

As with our use of the trailing slash in $MESSAGE_ID/T/ and
'$MESSAGE_ID/t/' endpoints, this for 'wget -r --mirror'
compatibility as well as allowing sysadmins to quickly stand up
a static directory with "index.html" in it to reduce load.

5 years agosolver: add a TODO note about making this fully evented
Eric Wong [Sat, 19 Jan 2019 08:27:44 +0000 (08:27 +0000)]
solver: add a TODO note about making this fully evented

Applying a 100+ patch series can be a pain and lead to a wayward
client monopolizing the connection.  On the other hand, we'll
also need to be careful and limit the number of in-flight file
descriptors and parallel git-apply processes when we move to an
evented model, here.

5 years agosolver: note the synchronous nature of index preparation
Eric Wong [Sat, 19 Jan 2019 08:13:11 +0000 (08:13 +0000)]
solver: note the synchronous nature of index preparation

It's not likely to be worth our time to support
a callback-driven model for something which happens
once per patch series.

5 years agosolver: break @todo loop into a callback
Eric Wong [Sat, 19 Jan 2019 06:37:53 +0000 (06:37 +0000)]
solver: break @todo loop into a callback

This will allow each patch search via Xapian to "yield" the
current client in favor of another client in the PSGI web
interface for fairness.

5 years agosolver: simplify control flow for initial loop
Eric Wong [Sat, 19 Jan 2019 05:44:12 +0000 (05:44 +0000)]
solver: simplify control flow for initial loop

We'll be breaking this up into several steps, too; since
searching inboxes for patch blobs can take 10s of milliseconds
for me.

5 years agosolver: switch patch application to use a callback
Eric Wong [Sat, 19 Jan 2019 05:25:30 +0000 (05:25 +0000)]
solver: switch patch application to use a callback

A bit messy at the moment, but we need to break this up
into smaller steps for fairness with other clients, as
applying dozens of patches can take several hundred
milliseconds.

5 years agosolver: break up patch application steps
Eric Wong [Sat, 19 Jan 2019 04:57:08 +0000 (04:57 +0000)]
solver: break up patch application steps

We want more fine-grained scheduling for PSGI use, as
the patch application step can take hundreds of milliseconds
on my modest hardware

5 years agosolver: more verbose blob resolution
Eric Wong [Sat, 19 Jan 2019 04:57:07 +0000 (04:57 +0000)]
solver: more verbose blob resolution

Help users find out where each step of the resolution came from.
Also, we must clean abort the process if we have missing blobs.

And refine the output to avoid unnecessary braces, too.

5 years agogit: support 'ambiguous' result from --batch-check
Eric Wong [Sat, 19 Jan 2019 06:36:04 +0000 (06:36 +0000)]
git: support 'ambiguous' result from --batch-check

David Turner's patch to return "ambiguous" seems like a reasonable
patch for future versions of git:

https://public-inbox.org/git/672a6fb9e480becbfcb5df23ae37193784811b6b.camel@novalis.org/

5 years agoview: enable naming hints for raw blob downloads
Eric Wong [Fri, 18 Jan 2019 12:18:10 +0000 (12:18 +0000)]
view: enable naming hints for raw blob downloads

Meaningful names in URLs are nice, and it can make
life easier for supporting syntax-highlighting

5 years agosolver: operate directly on git index
Eric Wong [Fri, 18 Jan 2019 11:17:58 +0000 (11:17 +0000)]
solver: operate directly on git index

No need to incur extra I/O traffic with a working-tree and
uncompressed files on the filesystem.  git can handle patch
application in memory and we rely on exact blob matching
anyways, so no need for 3way patch application.

5 years agogit: disable abbreviations with cat-file hints
Eric Wong [Fri, 18 Jan 2019 06:09:24 +0000 (06:09 +0000)]
git: disable abbreviations with cat-file hints

Ambiguity is not worth it for internal usage with the
solver.

5 years agoview: wire up diff and vcs viewers with solver
Eric Wong [Thu, 17 Jan 2019 11:51:34 +0000 (11:51 +0000)]
view: wire up diff and vcs viewers with solver

5 years agosolver: various bugfixes and cleanups
Eric Wong [Thu, 17 Jan 2019 11:50:57 +0000 (11:50 +0000)]
solver: various bugfixes and cleanups

Remove the make_path dependency and call mkdir directly.
Capture mode on new files, avoid referencing non-existent
functions and enhance the debug output for users to read.

5 years agogit: check saves error on disambiguation
Eric Wong [Fri, 18 Jan 2019 05:27:51 +0000 (05:27 +0000)]
git: check saves error on disambiguation

This will be useful for disambiguating short OIDs in older
emails when abbreviations were shorter.

Tested against the following script with /path/to/git.git
==> t.perl <==
use strict;
use PublicInbox::Git;
use Data::Dumper;
my $dir = shift or die "Usage: $0 GIT_DIR # (of git.git)";
my $git = PublicInbox::Git->new($dir);
my @res = $git->check('dead');
print Dumper({res => \@res, err=> $git->last_check_err});
@res = $git->check('5335669531d83d7d6c905bcfca9b5f8e182dc4d4');
print Dumper({res => \@res, err=> $git->last_check_err});

5 years agogit: add git_quote
Eric Wong [Fri, 18 Jan 2019 10:21:40 +0000 (10:21 +0000)]
git: add git_quote

It'll be helpful for displaying progress in SolverGit
output.

5 years agogit: support multiple URL endpoints
Eric Wong [Wed, 16 Jan 2019 21:53:07 +0000 (21:53 +0000)]
git: support multiple URL endpoints

For redundancy and centralization resistance.

5 years agosolver: initial Perl implementation
Eric Wong [Tue, 15 Jan 2019 08:22:41 +0000 (08:22 +0000)]
solver: initial Perl implementation

This will lookup git blobs from associated git source code
repositories.  If the blobs can't be found, an attempt to
"solve" them via patch application will be performed.

Eventually, this may become the basis of a type-agnostic
frontend similar to "git show"

5 years agot/perf-msgview: add test to check msg_html performance
Eric Wong [Tue, 15 Jan 2019 02:42:10 +0000 (02:42 +0000)]
t/perf-msgview: add test to check msg_html performance

This will be necessary to ensure we maintain reasonable
performance when we add diff-highlighting support.

5 years agohval: force monospace for <form> elements, too
Eric Wong [Sat, 19 Jan 2019 02:10:09 +0000 (02:10 +0000)]
hval: force monospace for <form> elements, too

Same reasoning as commit 7b7885fc3be2719c068c0a2fc860d53f17a1d933,
because GUI browsers have a tendency to use a different
font-family (and thus different size) as the rest of the page.

5 years agoview: disable bold in topic display
Eric Wong [Fri, 18 Jan 2019 06:19:09 +0000 (06:19 +0000)]
view: disable bold in topic display

It seems pointless due to the indentation, and interacts
badly with some CSS colouring.

5 years agot/git.t: do not pass "-b" to git-repack(1)
Eric Wong [Fri, 18 Jan 2019 21:43:29 +0000 (21:43 +0000)]
t/git.t: do not pass "-b" to git-repack(1)

Allows t/git.t to run on older versions of git without
"-b" and avoids incurring extra I/O traffic for bitmaps.

5 years agogit: git_unquote handles double-quote and backslash
Eric Wong [Fri, 18 Jan 2019 12:50:59 +0000 (12:50 +0000)]
git: git_unquote handles double-quote and backslash

We need to work with 0x22 (double-quote) and 0x5c (backslash);
even if they're oddball characters in filenames which wouldn't
be used by projects I'd want to work on.