Re: [gentoo-dev] Ebuild Help...

[Markus Brischke <zoke-gentoo@zoke.de>]

Quoting Matthew Walker (mwalker@kydance.net):
> I'm writing my first ebuild, and I'm having a hard time figuring out how
> to get it to install all the files properly once they've been created in
> the image directory. I've included my ebuild file. Any help would be
> greatly appreciated.
...
> src_install () {
>     make DESTDIR=${D} install || die
> 
>     dodir /usr/bin
>     dodir /usr/share/doc/${PN}-${PV}
>     dodir /usr/share/${PN}
>     dodir /usr/share/${PN}/images
>     dodir /usr/share/${PN}/sounds
> 
>     dobin ${D}/usr/bin/xscorch
>     dobin ${D}/usr/bin/xscorch-server
> 
>     insinto /usr/share/man/man6
>     doins ${D}/usr/share/man/man6/${PN}.6.gz
> 
>     dodoc AUTHORS ChangeLog COPYING INSTALL NEWS README TODO
> }

Hi, i have written yesterday my first too ;) If i read all the stuff
right the ${D} in the do<something> helperapps is unnecessary. This is
hardly ;) explained in the table in http://www.gentoo.org/doc/gentoo-howto.html
I read many of the ebuildfiles to get here and there a picture :) There
is a usefull manpage (man 5 ebuild) and a usefull skeleton and last but
not least the documentation and mailinglistarchive (is there a search
mechanism anywhere?).

If you are not sure if your installscript puts the thing into right
directories (meaning outside PORTAGE_TMPDIR) then use
FEATURES="sandbox". If i have read right this will protect you from
overwriting you system :) But (!) watch carefully imho not everything is
logged. e.g. the bernstein setup mechanism fails using mkdir(), but
there is no logwarning because no "new" file|dir created.

Another thing is the way how to install things. Imho it is a better
style to patch the Makefiles(.in) to improve or correct the ${D}
installation than to dodir/dobin/insinto,... doing yourself. The reason
for this the potential difference between a "make install" and doing the
"same" yourself. I found it maddening in some distros that the
packagemaintainers forgot these and that in the package that a "make
install" would have installed. Another adavantage would be if the author
of the package change the installation your patch won't properly apply
(in most cases not in all!). The Request for controlling the makefiles
is so right, but in these concepts even the controller make mistakes and
every hinting mechanism like sandox and (failed) patching should be used
but must not be relied on.

I hope this will help you and that i don't say anything wrong :)

Markus Brischke aka Zoke