Bugzilla Bug 63475

A new version of sys-apps/irda-utils is out. I've cleaned up the ebuild for
this release:

  * added gtk use flag for conditional compilation of gfindchip
  * corrected glib dependency
  * export WANT_AUTOMAKE not WANT_AUTOCONF
  * use emake instead of make
  * keep out of /usr/X11R6/bin

I have not added support for compiling the included proof-of-concept irnetd
(new in 0.9.16) since it I plan to provide an ebuild for the more advanced
IrNET daemon (which is written by Kevin Bertram) as a seperate ebuild.

Having irnetd as a seperate ebuild instead of the included one allows us to:
  * not add a dependency on net-dialup/ppp to sys-apps/irda-utils
  * avoid a conflict between sys-apps/irda-utils and the soon-to-come irnet
ebuild
  * have a more advanced irnetd

I also suggest we move the irda-utils ebuild to the net-wireless category which
is where it belongs. The metadata.xml lists base-system as the maintainer of
irda-utils - if possible I'd like to take over maintaining it and the upcoming
irnetd ebuild.

Let'me know if there are any objections.

Created an attachment (id=39271) [edit]
Patch for irda-utils-0.9.15.ebuild -> irda-utils-0.9.16.ebuild

On second thought - I don't think we would want to include neither the irnetd
from irda-utils nor the stand-alone irnetd.

Both implementations are rather 'hackish' and I can not get either to work. I
suggest we bump irda-utils to 0.9.16 using the attached ebuild and move it to
net-wireless.

Created an attachment (id=40203) [edit]
irda-utils-0.9.16.ebuild

Hi!

I had a few problems with irda-utils:
- with this ebuild I'm able to build irdadump succesfuly
- irda-utils compiles with gcc-3.4
- there's no {e}make install
- few more...

Patches comes from PLD Linux, Fedora, Mandrake and me :) .

I'm not sure about findchip (is it really x86 specific). I'll test it on my ppc
today / tommorow. (this info is from PLD Linux)

Hope you find it useful.

Regards,
Przemek

Created an attachment (id=40204) [edit]
irda-utils-gtk+2.patch

Created an attachment (id=40205) [edit]
irda-utils-includes.patch

Created an attachment (id=40206) [edit]
irda-utils-makefiles.patch

Created an attachment (id=40207) [edit]
irda-utils-gcc3.4-fix.patch

Created an attachment (id=40208) [edit]
irda-utils-tekram-ppc-buildfix.patch

Created an attachment (id=40209) [edit]
irda-utils-rh1.patch

Created an attachment (id=40210) [edit]
irda-utils-rootonly.patch

this is the only patch about one I'm not sure. 'security reasons'....?

Did you test attachment #39271 [edit] - the proposed patch for
irda-utils-0.9.16.ebuild - before attaching the modified ebuild and your
proposed patches?

Sorry - it's not that I do not now what this patch is doing... I'm just not
sure is it really necessery, but I think it may be useful.

And one more thing. In etc directory (in tarball file) there's modules.conf
example for 2.4 kernel. Maybe it would be ok to install this file in
/etc/modules.d directory and another one if kernel 2.6 is installed?

I wan't be able to do this now (work), but I'll do it tonight.

Regards,
Przemek

Yes - unfortunatelly:
- irdadump still wasn't build
- I had problems with gcc 3.4 (patch I submit is better that one created)
- ppc had problems with tekram
- there was now CFLAGS from /etc/make.conf for few programs (it's gentoo way - right?)
- on x86 I had problems with includes

With this ebuild evertyhing works as expected.

Regards,
Przemek

- I had problems with gcc 3.4 (patch I submit is better that one created)
- I had problems with gcc 3.4 (patch I submit is better that one I've created)

grrrrrr..... sorry for my english.

Reagrds,
Przemek

1. What problems have you encountered with the use of 'emake install'?

2. Porting gfindchip and friends from GTK+ to GTK2 is not as simple as s:AM_PATH_GLIB(1.2.0):PKG_CHECK_MODULES(GLIB, glib-2.0): - that patch can not be accepted.

3. Please attach the relevant error messages to justify the inclusion of the patches irda-utils-includes.patch, irda-utils-makefiles.patch, irda-utils-gcc3.4-fix.patch, irda-utils-tekram-ppc-buildfix.patch, irda-utils-rh1.patch, irda-utils-rootonly.patch

4. Has any of the above mentioned patches been submitted upstream?

5. Since the modules.conf entries rely on the particular hardware installed and the fact that irda-utils doesn't contain any kernel modules at all (these are provided by the kernel) I do not think we should install any default modules.conf file. If anything, we can install it to /usr/share/doc as an example configuration.

Please respond to the questions above so we can bump irda-utils.

Sorry for not responding to your post for such a long time. work, work, work...

1)
a) as i talk whive my friends from PLD Linux findchip is a bit broken, when not being used under x86 (atleast for now...). emake install is requiring findchip for an installation procces. to avoid compilation of findchip previously I've created patch to disable this (applied when ! use x86), but
b) during einstall I saw this:
make[1]: warning: jobserver unavailable: using -j1.  Add `+' to parent make rule

of course - this is not an error, but a simple warning. but it may look like a problem for users.

2) patch comes from PLD Linux. idea behind it is not to port, but to link whive a new library. since api haven't changed - why not?

ToBeContinuing.... in a minute

3)
irda-utils-makefiles.patch gives gentoo CFLAGS from /etc/make.conf for package during build procces. it's a gentoo way... right?

irda-utils-rootonly.patch is a paranoic security patch from fedora/mandrake . on one mailing list I saw a post, in which s/b said that in ealier version of irda-utils, there really was some security problem whive irdadump, but on.... windows ;-) as I said - this is paranoic sec. patch ;-)

3)
this is part of output on my ppc machine (gcc-3.4.1)

[cut]
In file included from /usr/src/linux/include/linux/byteorder/big_endian.h:11,
                 from /usr/src/linux/include/asm/byteorder.h:74,
                 from irkbd.c:48:
/usr/src/linux/include/linux/byteorder/swab.h:18:1: warning: "__attribute_const__" redefined
In file included from /usr/src/linux/include/linux/kd.h:4,
                 from /usr/include/sys/kd.h:27,
                 from irkbd.c:36:
/usr/src/linux/include/linux/compiler.h:120:1: warning: this is the location of the previous definition
irkbd.c: In function `irkbd_handle_response':
irkbd.c:289: error: parse error before string constant
make[1]: *** [irkbd.o] Error 1
[cut]
findchip.c:37:21: sys/io.h: No such file or directory
findchip.c: In function `main':
findchip.c:125: warning: implicit declaration of function `ioperm'
make[1]: *** [findchip.o] Error 1
[cut]

after applying:
irda-utils-includes.patch (kernel-2.6 issue?? only guessing...)
irda-utils-gcc3.4-fix.patch
irda-utils-tekram-ppc-buildfix.patch

everything is building as expected...

problem whive gcc3-4 is not only mine - friend from gentoo-forum-pl also had this issue on x86 machine. after applying those patches everything was fine.

3)
irda-utils-rh1.patch this patch is probably a comlite disaster... fedora and mandrake are adding it for a quite long time - and as I see only them. I have absolutely no idea why cutting a string is so important. i tried to google - but haven't find anything about this...

it is very probably that this patch isn't neccesery after all. I added it - mainly - since it's in fedora. it's not a bad distro (but it's rh based and not gentoo ;-) ).

4) don't know. sorry

5) well... i was thinking about file whive hashed lines. maybe it's a silly idea, but definitly user-friendly. this way a particular user only need ot unhash specific lines. only for this.

I think it's faster than gziping file in doc directory (for a user)

but 'dodoc' is gentoo way I suppose, so it's not a bad to :)

I hope my englisg isn't so bad and you can read it without much laugh ;-)

Regards,
Przemek

Hi,

Thank you for answering and clarifying things. I will look into applying these against irda-utils-0.9.16 and committing it to portage.

Regards,
Brix

I have checked the attached patches carefully - and I must admit that I find
several of them unacceptable for portage.

irda-utils-gtk+2.patch can not be accepted since we can not link with an API
incompatible version of the GTK+ library.

irda-utils-includes.patch can not be accepted since asm/io.h and sys/io.h is
not interchangeable.

irda-utils-makefiles.patch is not needed - a more simple approach is to export
RPM_OPT_FLAGS=${CFLAGS} before calling make

irda-utils-tekram-ppc-buildfix.patch can not be accepted since it's Gentoo
policy to build against kernel headers found in /usr/src/linux/include


irda-utils-gcc3.4-fix.patch is a valid fix for the deprecated __FUNCTION__
macro

irda-utils-rh1.patch is a valid fix for an off-by-one error in array index

Version 0.9.16 is now in portage. Closing as FIXED.