[Buildroot] libcurl patch for package/libcurl.mk

[Buildroot] libcurl patch for package/libcurl.mk

[Robert Urban]

apparently the boolean workaround (typedef unsigned char _Bool) gets in the way
of a native (gcc?) boolean datatype.  Below is a patch for libcurl.mk.

My buildroot version is

	BR2_VERSION="2010.05-rc2"

my host platform is Fedora-11, x86_64.

cheers,

Rob Urban

diff --git a/package/libcurl/libcurl.mk b/package/libcurl/libcurl.mk
index 863c301..f9e7b08 100644
--- a/package/libcurl/libcurl.mk
+++ b/package/libcurl/libcurl.mk
@@ -12,6 +12,7 @@ LIBCURL_CONF_OPT = --disable-verbose --disable-manual
--enable-hidden-symbols
 ifeq ($(BR2_PACKAGE_OPENSSL),y)
 LIBCURL_DEPENDENCIES += openssl
 LIBCURL_CONF_ENV += ac_cv_lib_crypto_CRYPTO_lock=yes
+LIBCURL_CONF_ENV += ac_cv_type_bool=yes
 # configure adds the cross openssl dir to LD_LIBRARY_PATH which screws up
 # native stuff during the rest of configure when target == host.
 # Fix it by setting LD_LIBRARY_PATH to something sensible so those libs

[Buildroot] libcurl patch for package/libcurl.mk

[Lionel Landwerlin]

Does this patch actually fix something ?
If it's just a matter of using something available in gcc, I would
rather prefer to let the configure script use its autodetect
feature....


On Fri, May 14, 2010 at 2:37 PM, Robert Urban <urban@unix-beratung.de> wrote:
> apparently the boolean workaround (typedef unsigned char _Bool) gets in the way
> of a native (gcc?) boolean datatype. ?Below is a patch for libcurl.mk.
>
> My buildroot version is
>
> ? ? ? ?BR2_VERSION="2010.05-rc2"
>
> my host platform is Fedora-11, x86_64.
>
> cheers,
>
> Rob Urban
>
> diff --git a/package/libcurl/libcurl.mk b/package/libcurl/libcurl.mk
> index 863c301..f9e7b08 100644
> --- a/package/libcurl/libcurl.mk
> +++ b/package/libcurl/libcurl.mk
> @@ -12,6 +12,7 @@ LIBCURL_CONF_OPT = --disable-verbose --disable-manual
> --enable-hidden-symbols
> ?ifeq ($(BR2_PACKAGE_OPENSSL),y)
> ?LIBCURL_DEPENDENCIES += openssl
> ?LIBCURL_CONF_ENV += ac_cv_lib_crypto_CRYPTO_lock=yes
> +LIBCURL_CONF_ENV += ac_cv_type_bool=yes
> ?# configure adds the cross openssl dir to LD_LIBRARY_PATH which screws up
> ?# native stuff during the rest of configure when target == host.
> ?# Fix it by setting LD_LIBRARY_PATH to something sensible so those libs
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot
>

[Buildroot] libcurl patch for package/libcurl.mk

[Lionel Landwerlin]

What was your problem ? Does the compilation failed ? Any log of the failure ?

I think we should find why the configure script fails to define
HAVE_BOOL_T, and try to fix the autodetect problem.

On Fri, May 14, 2010 at 3:11 PM, Robert Urban <urban@unix-beratung.de> wrote:
> Hi Lionel,
>
>
> On 05/14/2010 03:06 PM, Lionel Landwerlin wrote:
>> Does this patch actually fix something?
>
> it definately fixed it for me.
>
>> If it's just a matter of using something available in gcc, I would
>> rather prefer to let the configure script use its autodetect
>> feature....
>
> configure was getting something wrong. ?It wasn't defining HAVE_BOOL_T, but I
> don't know why.
>
> cheers,
>
> Rob Urban
>
>>
>> On Fri, May 14, 2010 at 2:37 PM, Robert Urban <urban@unix-beratung.de> wrote:
>>> apparently the boolean workaround (typedef unsigned char _Bool) gets in the way
>>> of a native (gcc?) boolean datatype. ?Below is a patch for libcurl.mk.
>>>
>>> My buildroot version is
>>>
>>> ? ? ? ?BR2_VERSION="2010.05-rc2"
>>>
>>> my host platform is Fedora-11, x86_64.
>>>
>>> cheers,
>>>
>>> Rob Urban
>>>
>>> diff --git a/package/libcurl/libcurl.mk b/package/libcurl/libcurl.mk
>>> index 863c301..f9e7b08 100644
>>> --- a/package/libcurl/libcurl.mk
>>> +++ b/package/libcurl/libcurl.mk
>>> @@ -12,6 +12,7 @@ LIBCURL_CONF_OPT = --disable-verbose --disable-manual
>>> --enable-hidden-symbols
>>> ?ifeq ($(BR2_PACKAGE_OPENSSL),y)
>>> ?LIBCURL_DEPENDENCIES += openssl
>>> ?LIBCURL_CONF_ENV += ac_cv_lib_crypto_CRYPTO_lock=yes
>>> +LIBCURL_CONF_ENV += ac_cv_type_bool=yes
>>> ?# configure adds the cross openssl dir to LD_LIBRARY_PATH which screws up
>>> ?# native stuff during the rest of configure when target == host.
>>> ?# Fix it by setting LD_LIBRARY_PATH to something sensible so those libs
>>> _______________________________________________
>>> buildroot mailing list
>>> buildroot at busybox.net
>>> http://lists.busybox.net/mailman/listinfo/buildroot
>>>
>>
>
>

[Buildroot] libcurl patch for package/libcurl.mk

[Lionel Landwerlin]

Le lundi 24 mai 2010 ? 15:11 +0200, Robert Urban a ?crit :
> Hi Lionel,
> 
> along with the libcurl problem, I had a series of other similar problems with
> packages whose builds failed in very confusing ways, for me at least.
> 
> Then I turned BR2_CONFIG_CACHE off, and the problems seem to have disappeared,
> including the libcurl problem. I'm not sure it was BR2_CONFIG_CACHE, because I
> changed other things as well, and got the latest versions from the repository,
> but BR2_CONFIG_CACHE does seem like the likely culprit.
> 
> Have you seen any problems with this?

No problem so far...

> 
> cheers,
> 
> Rob Urban
> 
> On 05/14/2010 03:19 PM, Lionel Landwerlin wrote:
> > What was your problem ? Does the compilation failed ? Any log of the failure ?
> > 
> > I think we should find why the configure script fails to define
> > HAVE_BOOL_T, and try to fix the autodetect problem.
> > 
> > On Fri, May 14, 2010 at 3:11 PM, Robert Urban <urban@unix-beratung.de> wrote:
> >> Hi Lionel,
> >>
> >>
> >> On 05/14/2010 03:06 PM, Lionel Landwerlin wrote:
> >>> Does this patch actually fix something?
> >>
> >> it definately fixed it for me.
> >>
> >>> If it's just a matter of using something available in gcc, I would
> >>> rather prefer to let the configure script use its autodetect
> >>> feature....
> >>
> >> configure was getting something wrong.  It wasn't defining HAVE_BOOL_T, but I
> >> don't know why.
> >>
> >> cheers,
> >>
> >> Rob Urban
> >>
> >>>
> >>> On Fri, May 14, 2010 at 2:37 PM, Robert Urban <urban@unix-beratung.de> wrote:
> >>>> apparently the boolean workaround (typedef unsigned char _Bool) gets in the way
> >>>> of a native (gcc?) boolean datatype.  Below is a patch for libcurl.mk.
> >>>>
> >>>> My buildroot version is
> >>>>
> >>>>        BR2_VERSION="2010.05-rc2"
> >>>>
> >>>> my host platform is Fedora-11, x86_64.
> >>>>
> >>>> cheers,
> >>>>
> >>>> Rob Urban
> >>>>
> >>>> diff --git a/package/libcurl/libcurl.mk b/package/libcurl/libcurl.mk
> >>>> index 863c301..f9e7b08 100644
> >>>> --- a/package/libcurl/libcurl.mk
> >>>> +++ b/package/libcurl/libcurl.mk
> >>>> @@ -12,6 +12,7 @@ LIBCURL_CONF_OPT = --disable-verbose --disable-manual
> >>>> --enable-hidden-symbols
> >>>>  ifeq ($(BR2_PACKAGE_OPENSSL),y)
> >>>>  LIBCURL_DEPENDENCIES += openssl
> >>>>  LIBCURL_CONF_ENV += ac_cv_lib_crypto_CRYPTO_lock=yes
> >>>> +LIBCURL_CONF_ENV += ac_cv_type_bool=yes
> >>>>  # configure adds the cross openssl dir to LD_LIBRARY_PATH which screws up
> >>>>  # native stuff during the rest of configure when target == host.
> >>>>  # Fix it by setting LD_LIBRARY_PATH to something sensible so those libs
> >>>> _______________________________________________
> >>>> buildroot mailing list
> >>>> buildroot at busybox.net
> >>>> http://lists.busybox.net/mailman/listinfo/buildroot
> >>>>
> >>>
> >>
> >>
> > 
> 

[Buildroot] libcurl patch for package/libcurl.mk

[Peter Korsgaard]

>>>>> "Lionel" == Lionel Landwerlin <llandwerlin@gmail.com> writes:

 Lionel> What was your problem ? Does the compilation failed ? Any log of the failure ?
 Lionel> I think we should find why the configure script fails to define
 Lionel> HAVE_BOOL_T, and try to fix the autodetect problem.

Indeed. I had a look at it today, and was able to reproduce it with
iperf. I have committed a workaround for it in iperf.mk.

-- 
Bye, Peter Korsgaard