Re: CVS commit: src/usr.bin/mail

[Steffen "Daode" Nurpmeso <sdaoden%gmail.com@localhost>]

christos%astron.com@localhost (Christos Zoulas) wrote:
 |In article <20130307104716.fxf5kembQ0XOuLi3FEqsN/NG@dietcurd.local>,
 |Steffen "Daode" Nurpmeso  <sdaoden%gmail.com@localhost> wrote:
 |>"Christos Zoulas" <christos%netbsd.org@localhost> wrote:
 |>|Modified Files:
 |>|    src/usr.bin/mail: fio.c
 |>|
 |>|Log Message:
 |>|since we are calling realpath() earlier now, we need to expand the name
 |>|of the folder in case it contained variables like $HOME.
 |>
 |>$HOME would simply be
 |>
 |>  :set folder=
 |
 |Yes, and how do you get:
 |
 |set MBOX=$HOME/Mail
 |set record=$HOME/Mail/sent-items

Not at all;  so it's a plain improvement of the codebase.
How about caching the result, just as S-nail does; i think with
your change you fork a shell subprocess for each and every
getfold() now?

 |>however; Mail(1) (and POSIX) yet don't document expansion, but
 |>only state
 |>
 |>  If this name begins with a `/', mail considers it to be an
 |>  absolute pathname; otherwise, the folder directory is found
 |>  relative to your home directory.
 |
 |Yes, but this worked and broke after your changes.

No no no no, we both know that you are going recursive:

  :set folder=+
  :fi

I'm just a dust-dry German from somewhere in the industrial
estate, but how happily i would twitter like a canary ...
Hah!  Hey Mister, your recent commit just broke software!
But i have a gut feeling .. and know some birds that sound much
sweeter than that, in addition :)

 |>patch in PR bin/47577 (the clean one using mblen(3))?
 |>It's working fine for me on UTF-8, and should do well for all
 |>multibyte encodings that have a clear notion of sequence leaders
 |>and follow bytes ('don't know wether there are ambiguous ones);
 |>it is anyway better than the version in CVS in respect to that.
 |
 |I really don't want to do a one-off patch for partial handling of
 |multi-byte characters. If we are going to support multi-byte characters
 |it should work everywhere.

The messed up patch happened just because i first wanted to use
mbtowc() before i realized that mblen(3) exists and is declared in
stdlib.h; sorry, sorry.  I hate this entire interface (and it's
not really working, since it cannot handle character sequences as
opposed to character codepoints anyway), and just never used it
myself.

I am the opinion that users should use paths in their native
language when they want to, and this may look odd with Mail(1)
without this patch.  And it must start somewhere.
And true: the patch also cleans up the code flo.

 |christos

Funky monk cheerio!

--steffen