[OpenWrt-Devel] HELP : Can't patch config.sub in a package

2012-01-22 Thread LEO Airwarosu Yoichi Shinoda
Hello,

I'm trying to add a new processor type (lexra), but having
trouble patching the config.sub, that checks for recognizable
cpu-os-kernel identifier, used in some packages.
More specifically, I want the config.sub in iptables-1.4.10 to
recognize the 'lexra-openwrt-linux'.

I'm working on the trunk r29839.

In the first place, I thought I can do this as a piece of cake
by adding a patch file that adds couple of lines to the config.sub,
in ${BUILDROOT}/package/iptables/patches.

Then, after few attempts, I found out that the autoreconf or a similar
mechanism replaces
${BUILDROOT}/build_dir/${TARGETSPEC}/iptables-1.4.10/config.sub
during the build process so that it points (symlinks) to
${BUILDROOT}/staging_dir/host/share/automake-1.11/config.sub.

At this point, I turned around and added a similar patch to
${BUILDROOT}/tools/automake/patches.

I checked my result immediately after the automake build,
and confirmed that the config.sub was patched appropriately.

But, ...the next make V=99 only revealed that the
${BUILDROOT}/staging_dir/host/share/automake-1.11/config.sub
was magically overwritten to the original unpatched state.

Here's my question:
I'm not an expert on OpenWrt build system, and therefore
I believe this is some kind of magic, and I also believe that
I need a same kind of magic to overcome this problem.
Is there any?

Thanks in advance,

--- shinoda

P.S.
Here's some discussion:
Is the behavior of Modifying a generic build tool during a build
process for a single target considered legitimate?
Wouldn't it introduce some unexpected side effects?

___
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel


Re: [OpenWrt-Devel] HELP : Can't patch config.sub in a package

2012-01-22 Thread Jo-Philipp Wich
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

Hi.


 I need a same kind of magic to overcome this problem.
 Is there any?

Yes, patch scripts/config.sub - that is whats fored upon each package.

 Is the behavior of Modifying a generic build tool during a build
 process for a single target considered legitimate?

Modifying the generic build tool to accept a new platform is certainly
legitimate, nobody wants to patch a new CPU definition into each single
package.

 Wouldn't it introduce some unexpected side effects?

Its autohell so nobody can tell :)


~ Jow
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk8cVnkACgkQdputYINPTPOmHQCeK+HDFGdXAfX8QjM4trm9B1uM
RDoAoIB1Gu3rRoaRs3OVkviYxsET6qLx
=7I17
-END PGP SIGNATURE-
___
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel


Re: [OpenWrt-Devel] HELP : Can't patch config.sub in a package

2012-01-22 Thread LEO Airwarosu Yoichi Shinoda

Jow,

 Yes, patch scripts/config.sub - that is whats fored upon each package.

Yikes!!!

Thanks a billion!!! I've been caught in this trap for half a day.

Packages were built flawlessly.

 Is the behavior of Modifying a generic build tool during a build
 process for a single target considered legitimate?
 
 Modifying the generic build tool to accept a new platform is certainly
 legitimate, nobody wants to patch a new CPU definition into each single
 package.

Ah, then, automake's config.sub should be patched once when it is built? :-)
(There's no argument, just my personal thoughts.)

--- shinoda

On 2012/01/23, at 3:33, Jo-Philipp Wich wrote:

 -BEGIN PGP SIGNED MESSAGE-
 Hash: SHA1
 
 Hi.
 
 
 I need a same kind of magic to overcome this problem.
 Is there any?
 
 Yes, patch scripts/config.sub - that is whats fored upon each package.
 
 Is the behavior of Modifying a generic build tool during a build
 process for a single target considered legitimate?
 
 Modifying the generic build tool to accept a new platform is certainly
 legitimate, nobody wants to patch a new CPU definition into each single
 package.
 
 Wouldn't it introduce some unexpected side effects?
 
 Its autohell so nobody can tell :)
 
 
 ~ Jow
 -BEGIN PGP SIGNATURE-
 Version: GnuPG v1.4.10 (GNU/Linux)
 Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
 
 iEYEARECAAYFAk8cVnkACgkQdputYINPTPOmHQCeK+HDFGdXAfX8QjM4trm9B1uM
 RDoAoIB1Gu3rRoaRs3OVkviYxsET6qLx
 =7I17
 -END PGP SIGNATURE-
 ___
 openwrt-devel mailing list
 openwrt-devel@lists.openwrt.org
 https://lists.openwrt.org/mailman/listinfo/openwrt-devel

___
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel


Re: [OpenWrt-Devel] HELP : Can't patch config.sub in a package

2012-01-22 Thread Jo-Philipp Wich
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

Hi.

 Modifying the generic build tool to accept a new platform is certainly
 legitimate, nobody wants to patch a new CPU definition into each single
 package.

In general yes, in practice it should be unneeded though since we're
forced to replace config.sub anyway in each package since all ship
different versions of config.sub and not all of them are
autoreconfigurable.

See Build/configure/Default in include/package-defaults.mk, thats where
we replace config.sub with our central scripts/ version in each package.

~ Jow
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk8cXk4ACgkQdputYINPTPM3WgCfdVgW0f3ac8SKo/gmMKuFw87R
RgUAoIbQj3NaCsR4m1dgpT85uOTh5w4/
=dQWJ
-END PGP SIGNATURE-
___
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel