[1/1] bash: make job control really work

Submitted by Dexuan Cui on Nov. 1, 2011, 8:05 a.m.

Details

Message ID ac7b821f266e932e267a4d62af83586ba3261a98.1320134607.git.dexuan.cui@intel.com
State New, archived
Headers show

Commit Message

Dexuan Cui Nov. 1, 2011, 8:05 a.m.
It turns out 9393ff833f44570fd5f500bc9de6c72db94b0296 didn't really fix
the bug.

This patch is made and tested after I read the link below
http://www.mail-archive.com/bug-bash@gnu.org/msg03107.html

[YOCTO #487]

Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
---
 meta/recipes-extended/bash/bash.inc    |    1 +
 meta/recipes-extended/bash/bash_4.2.bb |    2 +-
 2 files changed, 2 insertions(+), 1 deletions(-)

Patch hide | download patch | download mbox

diff --git a/meta/recipes-extended/bash/bash.inc b/meta/recipes-extended/bash/bash.inc
index d55e517..d495538 100644
--- a/meta/recipes-extended/bash/bash.inc
+++ b/meta/recipes-extended/bash/bash.inc
@@ -23,6 +23,7 @@  ALTERNATIVE_LINK = "${base_bindir}/sh"
 ALTERNATIVE_PRIORITY = "100"
 
 do_configure () {
+	export bash_cv_job_control_missing=present
 	gnu-configize
 	oe_runconf
 }
diff --git a/meta/recipes-extended/bash/bash_4.2.bb b/meta/recipes-extended/bash/bash_4.2.bb
index a0f5e4e..8070918 100644
--- a/meta/recipes-extended/bash/bash_4.2.bb
+++ b/meta/recipes-extended/bash/bash_4.2.bb
@@ -1,6 +1,6 @@ 
 require bash.inc
 
-PR = "r0"
+PR = "r1"
 
 SRC_URI = "${GNU_MIRROR}/bash/${BPN}-${PV}.tar.gz;name=tarball \
            ${GNU_MIRROR}/bash/bash-4.2-patches/bash42-001;apply=yes;striplevel=0;name=patch001 \

Comments

Richard Purdie Nov. 1, 2011, 10:21 a.m.
On Tue, 2011-11-01 at 16:05 +0800, Dexuan Cui wrote:
> It turns out 9393ff833f44570fd5f500bc9de6c72db94b0296 didn't really fix
> the bug.
> 
> This patch is made and tested after I read the link below
> http://www.mail-archive.com/bug-bash@gnu.org/msg03107.html
> 
> [YOCTO #487]
> 
> Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
> ---
>  meta/recipes-extended/bash/bash.inc    |    1 +
>  meta/recipes-extended/bash/bash_4.2.bb |    2 +-
>  2 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/meta/recipes-extended/bash/bash.inc b/meta/recipes-extended/bash/bash.inc
> index d55e517..d495538 100644
> --- a/meta/recipes-extended/bash/bash.inc
> +++ b/meta/recipes-extended/bash/bash.inc
> @@ -23,6 +23,7 @@ ALTERNATIVE_LINK = "${base_bindir}/sh"
>  ALTERNATIVE_PRIORITY = "100"
>  
>  do_configure () {
> +	export bash_cv_job_control_missing=present
>  	gnu-configize
>  	oe_runconf
>  }

This really should go into the common site files...

Cheers,

Richard
Dexuan Cui Nov. 2, 2011, 7:10 a.m.
Richard Purdie wrote on 2011-11-01:
> On Tue, 2011-11-01 at 16:05 +0800, Dexuan Cui wrote:
>> It turns out 9393ff833f44570fd5f500bc9de6c72db94b0296 didn't really
>> fix the bug.
>> 
>> This patch is made and tested after I read the link below
>> http://www.mail-archive.com/bug-bash@gnu.org/msg03107.html
>> 
>> [YOCTO #487]
>> 
>> Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
>> ---
>>  meta/recipes-extended/bash/bash.inc    |    1 +
>>  meta/recipes-extended/bash/bash_4.2.bb |    2 +-
>>  2 files changed, 2 insertions(+), 1 deletions(-)
>> diff --git a/meta/recipes-extended/bash/bash.inc
>> b/meta/recipes-extended/bash/bash.inc
>> index d55e517..d495538 100644
>> --- a/meta/recipes-extended/bash/bash.inc
>> +++ b/meta/recipes-extended/bash/bash.inc
>> @@ -23,6 +23,7 @@ ALTERNATIVE_LINK = "${base_bindir}/sh"
>>  ALTERNATIVE_PRIORITY = "100"
>>  
>>  do_configure () { +	export bash_cv_job_control_missing=present
>>  	gnu-configize 	oe_runconf }
> 
> This really should go into the common site files...
Hi Richard,
I found we do define the variable:
meta/site/common-linux:33:bash_cv_job_control_missing=${bash_cv_job_control_missing=present}
but looks autoconf doesn't realize the variable has been assigned the value 'present'?
I think this is because of the below do_configure in bash.inc -- looks autoreconf is skipped?
do_configure () {
    gnu-configize
    oe_runconf
}
Why do we need a customized do_configure to replace autotools_do_configure?

Later, after I added
do_configure_prepend () {
    autoreconf -f -i -s
}
The generated config.log does show bash_cv_job_control_missing is assigned with 'present'.
(BTW, common-linux also introduces many other variables -- would this be safe? Actually here I only need to introduce bash_cv_job_control_missing.)

However, finally, do_compile got a strange failure:
| shell.c: In function 'shell_reinitialize':
| shell.c:1742:20: error: 'PPROMPT' undeclared (first use in this function)
| shell.c:1742:20: note: each undeclared identifier is reported only once for each function it appears in
| shell.c:1743:22: error: 'SPROMPT' undeclared (first use in this function)

Could you please give some suggestions?

Thanks,
-- Dexuan
Richard Purdie Nov. 2, 2011, 10:35 a.m.
On Wed, 2011-11-02 at 15:10 +0800, Cui, Dexuan wrote:
> Richard Purdie wrote on 2011-11-01:
> > On Tue, 2011-11-01 at 16:05 +0800, Dexuan Cui wrote:
> >> It turns out 9393ff833f44570fd5f500bc9de6c72db94b0296 didn't really
> >> fix the bug.
> >> 
> >> This patch is made and tested after I read the link below
> >> http://www.mail-archive.com/bug-bash@gnu.org/msg03107.html
> >> 
> >> [YOCTO #487]
> >> 
> >> Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
> >> ---
> >>  meta/recipes-extended/bash/bash.inc    |    1 +
> >>  meta/recipes-extended/bash/bash_4.2.bb |    2 +-
> >>  2 files changed, 2 insertions(+), 1 deletions(-)
> >> diff --git a/meta/recipes-extended/bash/bash.inc
> >> b/meta/recipes-extended/bash/bash.inc
> >> index d55e517..d495538 100644
> >> --- a/meta/recipes-extended/bash/bash.inc
> >> +++ b/meta/recipes-extended/bash/bash.inc
> >> @@ -23,6 +23,7 @@ ALTERNATIVE_LINK = "${base_bindir}/sh"
> >>  ALTERNATIVE_PRIORITY = "100"
> >>  
> >>  do_configure () { +	export bash_cv_job_control_missing=present
> >>  	gnu-configize 	oe_runconf }
> > 
> > This really should go into the common site files...
> Hi Richard,
> I found we do define the variable:
> meta/site/common-linux:33:bash_cv_job_control_missing=${bash_cv_job_control_missing=present}
> but looks autoconf doesn't realize the variable has been assigned the value 'present'?
> I think this is because of the below do_configure in bash.inc -- looks autoreconf is skipped?
> do_configure () {
>     gnu-configize
>     oe_runconf
> }
> Why do we need a customized do_configure to replace autotools_do_configure?
> 
> Later, after I added
> do_configure_prepend () {
>     autoreconf -f -i -s
> }
> The generated config.log does show bash_cv_job_control_missing is assigned with 'present'.
> (BTW, common-linux also introduces many other variables -- would this be safe? Actually here I only need to introduce bash_cv_job_control_missing.)
> 
> However, finally, do_compile got a strange failure:
> | shell.c: In function 'shell_reinitialize':
> | shell.c:1742:20: error: 'PPROMPT' undeclared (first use in this function)
> | shell.c:1742:20: note: each undeclared identifier is reported only once for each function it appears in
> | shell.c:1743:22: error: 'SPROMPT' undeclared (first use in this function)
> 
> Could you please give some suggestions?

This is why its a really bad idea for recipes to have their own
configure rather than using our core one :/.

I had a go at this problem myself and it took a bit of figuring out. The
problem is that bash ships "config.h.in" but autoheader overwrites it.
This removes the start/end includes from config.h or config-bot.h and
config-top.h. We can do something like this:

export AUTOHEADER = "true"

do_configure_prepend () {
       if [ ! -e acinclude.m4 ]; then
               cat aclocal.m4 > acinclude.m4
       fi
}
 
instead of the current do_configure override. The _prepend ensures the
bash specific macros are preserved and the export AUTOHEADER stops
autoheader from running at all.

Could you test those changes against bash 4.x and bash 3.x (replacing
the current do_configure in bash.inc) and then if that works send a
patch please?

Cheers,

Richard
Dexuan Cui Nov. 2, 2011, 3:48 p.m.
Richard Purdie wrote on 2011-11-02:
> On Wed, 2011-11-02 at 15:10 +0800, Cui, Dexuan wrote:
>> Richard Purdie wrote on 2011-11-01:
>>> On Tue, 2011-11-01 at 16:05 +0800, Dexuan Cui wrote:
> I had a go at this problem myself and it took a bit of figuring out.
> The problem is that bash ships "config.h.in" but autoheader overwrites it.
> This removes the start/end includes from config.h or config-bot.h and
> config-top.h. We can do something like this:
> 
> export AUTOHEADER = "true"
> 
> do_configure_prepend () {
>        if [ ! -e acinclude.m4 ]; then
>                cat aclocal.m4 > acinclude.m4
>        fi
> }
> 
> instead of the current do_configure override. The _prepend ensures the
> bash specific macros are preserved and the export AUTOHEADER stops
> autoheader from running at all.
> 
> Could you test those changes against bash 4.x and bash 3.x (replacing
> the current do_configure in bash.inc) and then if that works send a patch please?
Looks bug 487 only happens to bash 4.x and bash 3.x doesn't have such a bug (the
recipe of bash 3.x doesn't use bash.inc, either)

RP, thanks a lot for the help! Here is the new patch:
http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=dcui/bug487&id=b524337ed5670de2c0d294c3f6970e24f23847eb

Thanks,
-- Dexuan