Re: [PATCH v2 00/20] insert command

2012-11-25 Thread Mark Walters

This is series is looking good. I have a few minor issues but no more.

There are two parts I don't really know enough about to be certain
about: the fsyncing and the sigtrap bit. 

For fsync: I have some recollection that you have to fsync all subpaths
to the root to guarantee that it makes it to the disk. 

One in particular that may need more fsyncing is if you create a folder
A/B/C/D/E/ for the messages then I think you only fsync the message and
A/B/C/D/E/new.

My other comments are all very minor and are made in individual replies.

Best wishes

Mark


On Sun, 25 Nov 2012, Peter Wang noval...@gmail.com wrote:
 This series mainly addresses the issues raised by Mark:

 - check talloc failures
 - deadlock in maildir_open_tmp
 - stricter file modes (0600 and 0700)
 - shared tag operation parser with notmuch-tag.c
 - simplified mkdir_parents
 - trap SIGINT
 - fsync after writing and rename
 - added a couple of tests
 - man page wording
 - comments

 Due to new restriction on tags beginning with '-', an argument beginning with
 -- is no longer ambiguous so I have removed the optional -- separator
 between options and tag operations.

 Peter Wang (20):
   tag: factor out tag operation parsing
   tag: make tag operation parser public
   cli: add stub for insert command
   insert: open Maildir tmp file
   insert: copy stdin to Maildir tmp file
   insert: move file from Maildir tmp to new
   insert: add new message to database
   insert: support --folder option
   insert: prevent writes outside Maildir hierarchy
   insert: apply default tags to new message
   insert: parse command-line tag operations
   insert: apply command-line tag operations
   insert: add --create-folder option
   insert: fsync after writing tmp file
   insert: fsync new directory after rename
   insert: trap SIGINT and clean up
   insert: add copyright line from notmuch-deliver
   test: add tests for insert
   man: document 'insert' command
   man: reference notmuch-insert.1

  Makefile.local  |   1 +
  man/Makefile.local  |   1 +
  man/man1/notmuch-config.1   |   4 +-
  man/man1/notmuch-count.1|   4 +-
  man/man1/notmuch-dump.1 |   4 +-
  man/man1/notmuch-insert.1   |  60 +
  man/man1/notmuch-new.1  |   4 +-
  man/man1/notmuch-reply.1|   3 +-
  man/man1/notmuch-restore.1  |   3 +-
  man/man1/notmuch-search.1   |   3 +-
  man/man1/notmuch-show.1 |   3 +-
  man/man1/notmuch-tag.1  |   3 +-
  man/man1/notmuch.1  |   3 +-
  man/man5/notmuch-hooks.5|   4 +-
  man/man7/notmuch-search-terms.7 |   3 +-
  notmuch-client.h|  12 +
  notmuch-insert.c| 493 
 
  notmuch-tag.c   |  79 ---
  notmuch.c   |   3 +
  test/insert |  93 
  test/notmuch-test   |   1 +
  21 files changed, 733 insertions(+), 51 deletions(-)
  create mode 100644 man/man1/notmuch-insert.1
  create mode 100644 notmuch-insert.c
  create mode 100755 test/insert

 -- 
 1.7.12.1

 ___
 notmuch mailing list
 notmuch@notmuchmail.org
 http://notmuchmail.org/mailman/listinfo/notmuch
___
notmuch mailing list
notmuch@notmuchmail.org
http://notmuchmail.org/mailman/listinfo/notmuch


Re: [PATCH v2 00/20] insert command

2012-11-25 Thread David Bremner
Peter Wang noval...@gmail.com writes:

 - shared tag operation parser with notmuch-tag.

Sharing tag operations with notmuch-tag is definitely a good idea, but
it will also conflict with the changes of the series at 

   id:1353792017-31459-1-git-send-email-da...@tethera.net

Although I might be biased, my current plan is to merge that series
first, since it fixes a long standing bug in the dump/restore commands.
If anything about series seems particularly problematic for
notmuch-insert, let me know.  Of course we can always make changes
afterwards, there are no public API changes (i.e. stuff ./lib) in play
here.

d
___
notmuch mailing list
notmuch@notmuchmail.org
http://notmuchmail.org/mailman/listinfo/notmuch