[Buildroot] [PATCH 1/1] package/ecryptfs-utils: Add support without gettext

[Yann E. MORIN]

Vadim, All,

On 2019-03-13 21:15 +0200, Vadym Kochan spake thusly:
> On Wed, Mar 13, 2019 at 07:53:50PM +0100, Yann E. MORIN wrote:
> > On 2019-03-13 00:20 +0200, Vadim Kochan spake thusly:
[--SNIP--]
> > > +gettext() {
> > > +	if [ -z "${GETTEXT_PROG}" ]; then
> > > +		GETTEXT_PROG=$(which gettext)
> > > +		if [ $? != 0 ]; then
> > 
> > != is a string comparison. For numbers, you will want to use -ne (or
> > -eq), see below.
> > 
> > > +			GETTEXT_PROG=gettext_echo
> > > +		fi
> > > +	fi
> > > +
> > > +	${GETTEXT_PROG} "$@"
> > > +}
> > 
> > This means that a script that calls gettext more than once will actually
> > try to resolve it each time. This is not optimum.
> 
> But I check the GETTEXT_PROG variable on the start, if it is not empty then just:
> 	${GETTEXT_PROG} "$@"

Hmm... Right, and the resolution is indeed done only once. My bad.

> > Instead, I think ecryptfs-common should basically look something like:
> > 
> >     if ! which gettext >/dev/null 2>&1; then
> >         gettext() {
> >             if [ -n "${GETTEXT}" ]; then
> >                 # Weird construct so that script that are 'set -e'
> >                 # fail at the call site of gettext and not here.
> >                 "${GETTEXT}" "${@}" || return $?
> >                 return 0
> >             fi
> >             while [ ${#} -ne 0 ]; do
> >                 case "${1}" in
> >                   (-h)              printf "no help\n"; return 0;;
> >                   (-V)              printf "0.0.0\n"; return 0;;
> >                   (-d|--domain)     shift 2;;
> >                   (-d*|--domain=*)  shift 1;;
> >                   (-e|-E|-n)        shift 1;;
> >                   (-s)              shift 1;;  # Ignore?
> >                   (-*)              printf "invalid option '%s'\n" "${1}" >&2; return 1;;
> >                   (*)               break;;
> >                 esac
> >             done
> >             case ${#} in
> >               (0)   printf "missing arguments\n" >&2; return 1;;
> >               (1)   printf "%s" "${1}";;
> >               (2)   shift; printf "%s" "${2}";;
> >               (*)   printf "too many arguments\n" >&2; return 1;;
> >             esac
> >         }
> >     fi
> > 
> > I.e. if there is no gettext program, you define a function named
> > gettext.

The difference then between your code and mine, is that yours is doing a
lazy resolution (i.e. resolving at first call) while mine is doing an
early resolution (i.e. resolving when parsing).

Additionally, since your code is using a two-level function call
(gettext -> gettext_echo), it's a bit more complex to handle scritps
that are 'set -e' and ensure they get the error at the call site rather
than in internal details of the emulating function.

But my suggestion would be that we just provide a gettext wrapper script
(not function) when gettext is not installed. This is much simpler
because there is thus no need to patch upstream.

> Thanks for your comments, anyway conceptually this patch is wrong. I
> tried to do it in the right way - install ecryptfs-common from
> ecryptfs-utils in src/utils/Makefile.am, but faced with autoreconf issue
> (autoreconf is required because src/utils/Makefile.am is changed by
> patch which adds installation of ecryptfs-common), but autoreconf fails
> to regenerate needed files because it removes aclocal.m4 which has
> definition of AM_GLIB_GNU_GETTEXT macro (which is not defined by host-gettext

Dang, indeed it is a big hammer to add a dependency onto host-libglib2
just for this oe macro, which by the way is deprecated, and that
upstream suggests to remove in favour of the official gettext one:

    https://gitlab.gnome.org/GNOME/glib/blob/master/m4macros/glib-gettext.m4#L356
.to:
    https://gitlab.gnome.org/GNOME/glib/blob/master/m4macros/glib-gettext.m4#L437

That's a shame, as ecrypt-utils does noit even have a dependency on
libglib2 at all, except for this one macro. :-(

So, that could probably be a hint to fix that upstream in a more
reasonable fashion anyway?

Regards,
Yann E. MORIN.

> nowhere in output/host) but this macro is required by configure.ac. So,
> I am not sure I can provide conceptually better patch for now (but Yann's comments),
> may be it is possible to regenerate only src/utils/Makefile.in (from
> *.am file) instead of calling autoreconf for all *.am, *.ac files.
> 
> Regards,
> Vadim Kochan

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 561 099 427 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'