Re: How to prepare an ISO 9660 CD for booting via GRUB ?

["Vladimir 'φ-coder/phcoder' Serbinenko" <phcoder@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 3553 bytes --]

Thomas Schmitt wrote:
> Hi,
>
> many thanks for testing xorriso in new
> situations. 
>
>
>   
>> 1) It tries to build even if no iconv is present
>>     
>
> The tests in configure.ac resp. acinclude.m4
> are too lax, obviously. They aim for finding
> out whether one needs -liconv or not. But they
> don't assure that iconv() works.
> I will have to make experiments with a
> non-existing dummy function, as it seems
> impossible to disable iconv() on Linux.
>
>   
I think what you should check for is iconv.h since this missing header
caused compilation problem. Moreover it covers the case of libiconv w/o
libiconv-devel.
>   
>> 2) I needed to explicitly add LIBS=-liconv to configure depstite
>> configure properly detecting the need of -liconv. In some commands
>> actually -liconv was double: one from configure and another from LIBS
>> but somewhere it's not propagated properly
>>     
>
> This sounds a bit like a bug in autotools.
>
> But in the log you sent, i only find them in
> pairs. Obviously one from LIBS= and one from
>   dnl GNU iconv has no function iconv() but libiconv() and a macro iconv()
>   AC_CHECK_LIB(iconv, libiconv, , )
>
> Are you sure that the lack of -liconv was not
> due to libiconv missing initially ?
>
>   
You're right it was some leftover in build directory. When I rebuild
from clean untar it worked fine.
>   
>> 3) A ton of warnings.
>>     
>
> The implementation of the isspace() family seems
> not to be liked by the compiler. Shrug.
>
> I found legitimate warnings about putting out
> uid_t and gid_t by printf formatter "%d".
> (Question is whether (int) is a suitable cast
>  in this case.)
>   
/usr/include/sys/cygwin.h:typedef unsigned long __uid32_t;
So you need to use %lu or %lx on cygwin.
> Then there is the intentional warning that no
> MMC system adapter is available for operating
> optical drives directly.
>   libburn/sg.c:42: warning: unused variable INTENTIONAL_COMPILER_WARNING
>   libburn/sg.c:39: warning: unused variable This_libburn_will_not_be_able_to_operate_on_real_CD_drives
>
> This does not hamper the GRUB use case. One will
> just need cdrecord for burning the CD, rather
> than letting xorriso do:
>   xorriso -as cdrecord -v dev=... blank=as_needed my_grub.iso
>
>
> ------------------------------------------------
>
> So mainly the LIBS=-liconv remains as build
> obstacle.
>
> I have changed the iconv detection code in
> configure.ac meanwhile (to care for the "const"
> ambiguity on FreeBSD). A new tarball
>   http://www.gnu.org/software/xorriso/xorriso-0.5.3.tar.gz
> is uploaded.
>
>   cat xorriso/xorriso_timestamp.h
> should say
>   #define Xorriso_timestamP "2010.04.10.180103"
> or later.
>
> Could you please try it on cygwin without LIBS=
> (but libiconv installed) and send me the log of
> the failing compile run ?
>
>   
I've tested with previous tarball. I'll retest on both systems with new one.
> If nothing else helps then i plan to set
>   LIBS="$LIBS -liconv"
> in a special cygwin case in acinclude.m4.
> (Do you know a reliable test pattern for cygwin ?
>  Like *-*-linux* for Linux.)
>
> For now the double -liconv options give me hope
> that it could work without that.
>
>
> Have a nice day :)
>
> Thomas
>
>
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/grub-devel
>
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 293 bytes --]