kernel menuconfig/ncurses patch in linux-yocto

kernel menuconfig/ncurses patch in linux-yocto

[Jacob Kroon]

Hi,
I'm trying to get the SDK to be able to run the kernel's "make
menuconfig" target
using nativesdk-ncurses from the SDK. Looking at

http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc

I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
in that patch looks weird:

    elif pkg-config --cflags ncurses 2>/dev/null; then
        echo '-DCURSES_LOC="<ncurses.h>"'
+   if [ x"$CROSS_CURSES_INC" != x ]; then
+   echo "$CROSS_CURSES_INC"
+   exit
+   fi
    elif [ -f /usr/include/ncursesw/curses.h ]; then
        echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'

(I had to do manual indentation with spaces in gmail)
Is the indentation or the logic incorrect ?

Regards Jacob

Re: kernel menuconfig/ncurses patch in linux-yocto

[Bruce Ashfield]

On 2016-07-20 4:19 PM, Jacob Kroon wrote:
> Hi,
> I'm trying to get the SDK to be able to run the kernel's "make
> menuconfig" target
> using nativesdk-ncurses from the SDK. Looking at
>
> http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc
>
> I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
> in that patch looks weird:
>
>     elif pkg-config --cflags ncurses 2>/dev/null; then
>         echo '-DCURSES_LOC="<ncurses.h>"'
> +   if [ x"$CROSS_CURSES_INC" != x ]; then
> +   echo "$CROSS_CURSES_INC"
> +   exit
> +   fi
>     elif [ -f /usr/include/ncursesw/curses.h ]; then
>         echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>
> (I had to do manual indentation with spaces in gmail)
> Is the indentation or the logic incorrect ?

In the commit itself, the indentation is fine.

That block of code is just dumping flags that are used in the
build. So in this case, it is correct. If the variable is
non empty, it is echoed and then processing exits.

Bruce

>
> Regards Jacob
>

Re: kernel menuconfig/ncurses patch in linux-yocto

[Jacob Kroon]

On Thu, Jul 21, 2016 at 5:22 AM, Bruce Ashfield
<bruce.ashfield@windriver.com> wrote:
> On 2016-07-20 4:19 PM, Jacob Kroon wrote:
>>
>> Hi,
>> I'm trying to get the SDK to be able to run the kernel's "make
>> menuconfig" target
>> using nativesdk-ncurses from the SDK. Looking at
>>
>>
>> http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc
>>
>> I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
>> in that patch looks weird:
>>
>>     elif pkg-config --cflags ncurses 2>/dev/null; then
>>         echo '-DCURSES_LOC="<ncurses.h>"'
>> +   if [ x"$CROSS_CURSES_INC" != x ]; then
>> +   echo "$CROSS_CURSES_INC"
>> +   exit
>> +   fi
>>     elif [ -f /usr/include/ncursesw/curses.h ]; then
>>         echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>>
>> (I had to do manual indentation with spaces in gmail)
>> Is the indentation or the logic incorrect ?
>
>
> In the commit itself, the indentation is fine.
>
> That block of code is just dumping flags that are used in the
> build. So in this case, it is correct. If the variable is
> non empty, it is echoed and then processing exits.

The check if CROSS_CURSES_INC is non-empty is only done if the preceeding
 "elif"-check is true, and not unconditionally as one would expect
judging by the indentation.

Re: kernel menuconfig/ncurses patch in linux-yocto

[Bruce Ashfield]

On 2016-07-21 02:35 AM, Jacob Kroon wrote:
> On Thu, Jul 21, 2016 at 5:22 AM, Bruce Ashfield
> <bruce.ashfield@windriver.com> wrote:
>> On 2016-07-20 4:19 PM, Jacob Kroon wrote:
>>>
>>> Hi,
>>> I'm trying to get the SDK to be able to run the kernel's "make
>>> menuconfig" target
>>> using nativesdk-ncurses from the SDK. Looking at
>>>
>>>
>>> http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc
>>>
>>> I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
>>> in that patch looks weird:
>>>
>>>      elif pkg-config --cflags ncurses 2>/dev/null; then
>>>          echo '-DCURSES_LOC="<ncurses.h>"'
>>> +   if [ x"$CROSS_CURSES_INC" != x ]; then
>>> +   echo "$CROSS_CURSES_INC"
>>> +   exit
>>> +   fi
>>>      elif [ -f /usr/include/ncursesw/curses.h ]; then
>>>          echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>>>
>>> (I had to do manual indentation with spaces in gmail)
>>> Is the indentation or the logic incorrect ?
>>
>>
>> In the commit itself, the indentation is fine.
>>
>> That block of code is just dumping flags that are used in the
>> build. So in this case, it is correct. If the variable is
>> non empty, it is echoed and then processing exits.
>
> The check if CROSS_CURSES_INC is non-empty is only done if the preceeding
>   "elif"-check is true, and not unconditionally as one would expect
> judging by the indentation.

Yep, that's the point. It goes to pkgconfig first, and then falls
back to that.

The indention is fine in the actual repository.

--- a/scripts/kconfig/lxdialog/check-lxdialog.sh
+++ b/scripts/kconfig/lxdialog/check-lxdialog.sh
@@ -4,6 +4,10 @@
  # What library to link
  ldflags()
  {
+       if [ "$CROSS_CURSES_LIB" != "" ]; then
+               echo "$CROSS_CURSES_LIB"
+               exit
+       fi
         pkg-config --libs ncursesw 2>/dev/null && exit
         pkg-config --libs ncurses 2>/dev/null && exit
         for ext in so a dll.a dylib ; do
@@ -25,6 +29,10 @@ ccflags()
                 echo '-DCURSES_LOC="<ncurses.h>" -DNCURSES_WIDECHAR=1'
         elif pkg-config --cflags ncurses 2>/dev/null; then
                 echo '-DCURSES_LOC="<ncurses.h>"'
+       if [ x"$CROSS_CURSES_INC" != x ]; then
+               echo "$CROSS_CURSES_INC"
+               exit
+       fi
         elif [ -f /usr/include/ncursesw/curses.h ]; then
                 echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
                 echo ' -DNCURSES_WIDECHAR=1'



Bruce

>

Re: kernel menuconfig/ncurses patch in linux-yocto

[Bruce Ashfield]

On 2016-07-21 08:58 AM, Bruce Ashfield wrote:
> On 2016-07-21 02:35 AM, Jacob Kroon wrote:
>> On Thu, Jul 21, 2016 at 5:22 AM, Bruce Ashfield
>> <bruce.ashfield@windriver.com> wrote:
>>> On 2016-07-20 4:19 PM, Jacob Kroon wrote:
>>>>
>>>> Hi,
>>>> I'm trying to get the SDK to be able to run the kernel's "make
>>>> menuconfig" target
>>>> using nativesdk-ncurses from the SDK. Looking at
>>>>
>>>>
>>>> http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc
>>>>
>>>>
>>>> I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
>>>> in that patch looks weird:
>>>>
>>>>      elif pkg-config --cflags ncurses 2>/dev/null; then
>>>>          echo '-DCURSES_LOC="<ncurses.h>"'
>>>> +   if [ x"$CROSS_CURSES_INC" != x ]; then
>>>> +   echo "$CROSS_CURSES_INC"
>>>> +   exit
>>>> +   fi
>>>>      elif [ -f /usr/include/ncursesw/curses.h ]; then
>>>>          echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>>>>
>>>> (I had to do manual indentation with spaces in gmail)
>>>> Is the indentation or the logic incorrect ?
>>>
>>>
>>> In the commit itself, the indentation is fine.
>>>
>>> That block of code is just dumping flags that are used in the
>>> build. So in this case, it is correct. If the variable is
>>> non empty, it is echoed and then processing exits.
>>
>> The check if CROSS_CURSES_INC is non-empty is only done if the preceeding
>>   "elif"-check is true, and not unconditionally as one would expect
>> judging by the indentation.
>
> Yep, that's the point. It goes to pkgconfig first, and then falls
> back to that.
>
> The indention is fine in the actual repository.

oh wait.

I had my coffee and I see now what you mean, I was checking
older repos, and when this merged into the 4.4 tree it does
look like the conditional is hosed.

I'll queue up a closer look at this later today.

Bruce

>
> --- a/scripts/kconfig/lxdialog/check-lxdialog.sh
> +++ b/scripts/kconfig/lxdialog/check-lxdialog.sh
> @@ -4,6 +4,10 @@
>   # What library to link
>   ldflags()
>   {
> +       if [ "$CROSS_CURSES_LIB" != "" ]; then
> +               echo "$CROSS_CURSES_LIB"
> +               exit
> +       fi
>          pkg-config --libs ncursesw 2>/dev/null && exit
>          pkg-config --libs ncurses 2>/dev/null && exit
>          for ext in so a dll.a dylib ; do
> @@ -25,6 +29,10 @@ ccflags()
>                  echo '-DCURSES_LOC="<ncurses.h>" -DNCURSES_WIDECHAR=1'
>          elif pkg-config --cflags ncurses 2>/dev/null; then
>                  echo '-DCURSES_LOC="<ncurses.h>"'
> +       if [ x"$CROSS_CURSES_INC" != x ]; then
> +               echo "$CROSS_CURSES_INC"
> +               exit
> +       fi
>          elif [ -f /usr/include/ncursesw/curses.h ]; then
>                  echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>                  echo ' -DNCURSES_WIDECHAR=1'
>
>
>
> Bruce
>
>>
>

Re: kernel menuconfig/ncurses patch in linux-yocto

[Jacob Kroon]

On Thu, Jul 21, 2016 at 3:06 PM, Bruce Ashfield
<bruce.ashfield@windriver.com> wrote:
> On 2016-07-21 08:58 AM, Bruce Ashfield wrote:
>>
>> On 2016-07-21 02:35 AM, Jacob Kroon wrote:
>>>
>>> On Thu, Jul 21, 2016 at 5:22 AM, Bruce Ashfield
>>> <bruce.ashfield@windriver.com> wrote:
>>>>
>>>> On 2016-07-20 4:19 PM, Jacob Kroon wrote:
>>>>>
>>>>>
>>>>> Hi,
>>>>> I'm trying to get the SDK to be able to run the kernel's "make
>>>>> menuconfig" target
>>>>> using nativesdk-ncurses from the SDK. Looking at
>>>>>
>>>>>
>>>>>
>>>>> http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc
>>>>>
>>>>>
>>>>> I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
>>>>> in that patch looks weird:
>>>>>
>>>>>      elif pkg-config --cflags ncurses 2>/dev/null; then
>>>>>          echo '-DCURSES_LOC="<ncurses.h>"'
>>>>> +   if [ x"$CROSS_CURSES_INC" != x ]; then
>>>>> +   echo "$CROSS_CURSES_INC"
>>>>> +   exit
>>>>> +   fi
>>>>>      elif [ -f /usr/include/ncursesw/curses.h ]; then
>>>>>          echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>>>>>
>>>>> (I had to do manual indentation with spaces in gmail)
>>>>> Is the indentation or the logic incorrect ?
>>>>
>>>>
>>>>
>>>> In the commit itself, the indentation is fine.
>>>>
>>>> That block of code is just dumping flags that are used in the
>>>> build. So in this case, it is correct. If the variable is
>>>> non empty, it is echoed and then processing exits.
>>>
>>>
>>> The check if CROSS_CURSES_INC is non-empty is only done if the preceeding
>>>   "elif"-check is true, and not unconditionally as one would expect
>>> judging by the indentation.
>>
>>
>> Yep, that's the point. It goes to pkgconfig first, and then falls
>> back to that.
>>
>> The indention is fine in the actual repository.
>
>
> oh wait.
>
> I had my coffee and I see now what you mean, I was checking
> older repos, and when this merged into the 4.4 tree it does
> look like the conditional is hosed.
>
> I'll queue up a closer look at this later today.
>
> Bruce
>

Ah, that explains it. Thank you for looking into it.

Jacob

Re: kernel menuconfig/ncurses patch in linux-yocto

[Bruce Ashfield]

On 2016-07-21 10:24 AM, Jacob Kroon wrote:
> On Thu, Jul 21, 2016 at 3:06 PM, Bruce Ashfield
> <bruce.ashfield@windriver.com> wrote:
>> On 2016-07-21 08:58 AM, Bruce Ashfield wrote:
>>>
>>> On 2016-07-21 02:35 AM, Jacob Kroon wrote:
>>>>
>>>> On Thu, Jul 21, 2016 at 5:22 AM, Bruce Ashfield
>>>> <bruce.ashfield@windriver.com> wrote:
>>>>>
>>>>> On 2016-07-20 4:19 PM, Jacob Kroon wrote:
>>>>>>
>>>>>>
>>>>>> Hi,
>>>>>> I'm trying to get the SDK to be able to run the kernel's "make
>>>>>> menuconfig" target
>>>>>> using nativesdk-ncurses from the SDK. Looking at
>>>>>>
>>>>>>
>>>>>>
>>>>>> http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-4.4/commit/scripts/kconfig/lxdialog/check-lxdialog.sh?h=standard/beaglebone&id=badf6fedf455958fe0ff3c060c8e3965ef6d80dc
>>>>>>
>>>>>>
>>>>>> I figured out I could pass CROSS_CURSES_[INC,LIB], but the second chunk
>>>>>> in that patch looks weird:
>>>>>>
>>>>>>       elif pkg-config --cflags ncurses 2>/dev/null; then
>>>>>>           echo '-DCURSES_LOC="<ncurses.h>"'
>>>>>> +   if [ x"$CROSS_CURSES_INC" != x ]; then
>>>>>> +   echo "$CROSS_CURSES_INC"
>>>>>> +   exit
>>>>>> +   fi
>>>>>>       elif [ -f /usr/include/ncursesw/curses.h ]; then
>>>>>>           echo '-I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"'
>>>>>>
>>>>>> (I had to do manual indentation with spaces in gmail)
>>>>>> Is the indentation or the logic incorrect ?
>>>>>
>>>>>
>>>>>
>>>>> In the commit itself, the indentation is fine.
>>>>>
>>>>> That block of code is just dumping flags that are used in the
>>>>> build. So in this case, it is correct. If the variable is
>>>>> non empty, it is echoed and then processing exits.
>>>>
>>>>
>>>> The check if CROSS_CURSES_INC is non-empty is only done if the preceeding
>>>>    "elif"-check is true, and not unconditionally as one would expect
>>>> judging by the indentation.
>>>
>>>
>>> Yep, that's the point. It goes to pkgconfig first, and then falls
>>> back to that.
>>>
>>> The indention is fine in the actual repository.
>>
>>
>> oh wait.
>>
>> I had my coffee and I see now what you mean, I was checking
>> older repos, and when this merged into the 4.4 tree it does
>> look like the conditional is hosed.
>>
>> I'll queue up a closer look at this later today.
>>
>> Bruce
>>
>
> Ah, that explains it. Thank you for looking into it.

Yup. The original patch (of 3.14 vintage) had migrated that chunk over
time to an invalid location.

I fixed the block in the 4.4 kernel, and have refreshed the path so
it will be fixed for the next linux-yocto kernel as well.

I've pushed the changes to the tree, and will follow up with SRCREV
bumps when my next change queue is ready.

Bruce

>
> Jacob
>