hi ketul
ketul: if you haven't been paying attention to it, you should read the mailing list thread on users@rtems.org about "Embedded board with good support out of the box?"
ketul
hi gedare
ketul
gedare: https://gist.github.com/ketul93/d717555951174a74c8b4 updated !
but in this code getting one fatal error. gedare can you help me to rectify it
→ DrJoel has joined
← DrJoel has quit (Changing host)
→ DrJoel has joined
ⓘ ChanServ set mode +o DrJoel
ketul
hi DrJoel
@DrJoel
hello! How's your code doing?
ketul
https://gist.github.com/ketul93/d717555951174a74c8b4
can you please review it DrJoel
@DrJoel
It will be quick. I have to leave in a few minutes for a demo.
beagleboneblack.h appears to be OK except some enums and struct members are indented with one space not two. Two is the rule.
I assume there are no tabs.
ketul
ya seen that problem
@DrJoel
gpio.c: The API describing Doxygen goes in the .h.
gpio.c: returning 1, 2 for errors isn't good form. Define them as #define constants or as an enumeration for GPIO_ERRORS or something
ketul
DrJoel: that is defined in gpio.h
all the macros
for errors and application is there in init.c
@DrJoel
Each value returned should be covered in the comments. Comment for rtems_gpio_configure_pin_digital_out() says 0 or 1 but returns 0, 1, and 2. That's the method I am talking about.
ketul
ohh k
@DrJoel
The methods below that are documented as 0/-1 but return 0/2. Defining the error names as macros is OK but document what it means when they are returned.
gpio.h: I see the error defining macros but you didn't use them. :)
And the comments don't describe that method X can return it.
ketul
yup didn't use all
ok that I will comment
used only "GPIO_SUCCESSFUL"
@DrJoel
Using them all in a particular method is not an issue if you used the entire set across all methods. Key is unique status per condition per method. So if method X returns error A, it has a unique interpretation. At least that's my view of good error codes. No guessing what it means
beng-nl and gedare can comment but we may want to already use a BSD 2-clause license for this code if you don't mind. Ask them.
ketul
ok sure
thanks DrJoel
@DrJoel
You don't have to have extern on methods in gpio.h and the detailed Doxygen in the C file should be in here. That's what shows up in the generated output.
Did we decide on a macro a BSP defines in bsp.h when it supports gpio?
init.c: if( should be "if (
Some of the if lines look long. Make sure < 80 columns.
Putting the return code in a variable and testing the variable would be better since it makes it easier/possible to print the status when debugging.
Obviously the test is BB specific at this point. See line 14.
We need to work on how to make this code generic so it just works without knowing a BSP. Likely this means bsp.h will have to include the bsp/beagleboneblack.h when configured for that variant and the test will ahve to check BSP_HAS_GPIO or whatever macro we decide to say means it supports GPIO
ketul
yup :)
ketul
but DrJoel different boards have different methods of accessing the gpio pins so at line 14 it is defined that it is using the macros for BBB.
ok got it
:)
so you are talking about adding one more abstraction layer
@DrJoel
Does that make two or three layers? I think two are sufficient. Application level and BSP level
ketul
ya probably 2 layers
@DrJoel
OK.. have to head to demo
have a great weekend
← DrJoel has quit (Quit: Do fish get thirsty?)
rtems-bot
[trac] TBR/UserManual/MinGW_Tools_for_Windows edited by gedare (Fix ftp links.) http://devel.rtems.org/wiki/TBR/UserManual/MinGW_Tools_for_Windows
rtems-bot
[trac] 2320/defect">#2320/defect (new) updated by gedare (The following search shows where the old links are still existing in the wiki: ...) http://devel.rtems.org/ticket/2320
[trac] 2320/defect">#2320/defect (new) updated by gedare (empty comment) http://devel.rtems.org/ticket/2320
rtems-bot
[git] b17fb17 by Sebastian Huber:  [Revert "tmtests/tm26: Avoid NULL pointer access"  There is no NULL pointer access.  Please note that _Thread_Get_executing() != executing variable in Low_task().  This reverts commit 5611839a7e2e371dd1f327c336c785095f634e55.]  http://devel.rtems.org/changeset/b17fb17/rtems
→ Davidbrcz has joined
← Davidbrcz has quit (Ping timeout: 256 seconds)
→ hesham has joined
← javamonn has quit (Remote host closed the connection)
→ javamonn has joined
beng-nl
hi all
@gedare
hello beng-nl
← zoso has quit (Remote host closed the connection)
rtems-bot
[trac] SOCIS/2015 edited by hermann19829 (empty comment) http://devel.rtems.org/wiki/SOCIS/2015
→ Davidbrcz has joined
← javamonn has quit (Remote host closed the connection)
→ javamonn has joined
← javamonn has quit (Read error: Connection reset by peer)
→ javamonn has joined
→ mumptai has joined
ketul
hello beng-nl
beng-nl
hi ketul
ketul
beng-nl: getting fatal error "fatal error: rtems/gpio.h: No such file or directory"
I added the header to makefile.am
@gedare
ketul: bootstrap -p && bootstrap
ketul
all done
@gedare
what did you add to Makefile.am?
ketul
gedare: ## GPIO API
include_rtems_HEADERS += include/rtems/gpio.h
ketul
even it is showing gpio.h in preinstall.am
@gedare
paste the preinstall.am entry here
ketul
ok let me
@gedare
is this the beagle/Makefile.am?
ketul
no its in cpukit/
@gedare
k
ketul
gedare: $(PROJECT_INCLUDE)/rtems/gpio.h: include/rtems/gpio.h $(PROJECT_INCLUDE)/rtems/$
        $(INSTALL_DATA) $< $(PROJECT_INCLUDE)/rtems/gpio.h
PREINSTALL_FILES += $(PROJECT_INCLUDE)/rtems/gpio.h
so I added #include <rtems/gpio.h> in my code
but giving fatal error
@gedare
oh
anytime you do bootstrap -p and then bootstrap, you basically have to re-run configure and make too
so in your build dir, rm -rf *
and re-run configure
it's a nasty build system, and motivates the waf effort.
beng-nl
hear hear :)
ketul
:) yes beng-nl
ketul
gedare: yes I followed all the steps from "./bootstrap; ./bootstrap -p" to "make"
@gedare
you have to do bootstrap -p first
ketul, make sure you ran bootstrap again after bootstrap -p, and then ran configure and make in a clean build directory.
ketul
ok gedare let me try it
@gedare
ketul: any luck? i'm about to head out for the weekend.
ketul
compilation going on
gedare: have a great weekend
ketul
gedare: yup done with that error, but getting another one that I will try to rectify
gedare: thanks
← gedare has quit (Ping timeout: 265 seconds)
ketul
beng-nl: :)
ketul
beng-nl: wrote "typedef struct gpio_pin_handle;" in gpio.h
and struct declared in gpio.c
ketul
hi gedare free for some time ?
← verm__ has quit (Read error: Connection reset by peer)
→ verm__ has joined
→ Davidbrcz has joined
@gedare
hi ketul what's up?
ketul
hi gedare : have a look at gpio.h @ https://gist.github.com/ketul93/d717555951174a74c8b4
ketul
getting error "../../../../../.././beagleboneblack/lib/include/rtems/gpio.h:43:49: error: unknown type name 'gpio_pin_handle'
 extern int rtems_gpio_configure_pin_digital_out(gpio_pin_handle *, unsigned int pin_number);"
though I declared "typedef struct gpio_pin_handle;"
@gedare
sure, the filename is wong for beaglebonenlack.h
also
you shouldn't declare the typedef twice, so that is a bit fishy
probably the struct definition should be in gpio.h and not beaglebonenlack.h
ketul
ya file name was typing mistake
but
we need to make the gpio.h generalized api
so abstraction need to be there
@gedare
i see
ketul
so can't define struct in gpio.h
so what would you suggest ?
@gedare
then you also shouldn't be dereferencing pointers to it
either you define a struct that is generic across all GPIO implementations
or you define a struct that the GPIO implementation fills in somehow with some specialization such as function pointers or similar
ok i'm starting to see the picture
you want to know how to define the interface in a generic way
ketul
struct definition is board specific
ya kind of
or is it possible to use void * ?
@gedare
then you should not make it a typedef, just forward declare 'struct gpio_pin_handle;'
ketul
ok
@gedare
and use struct gpio_pin_handle in the interface functions
i think that should work...
*thinks*
otherwise, you define the struct there, and you let the gpio implementation sort of over-ride it, like...
typedef struct {
... generic stuff
} gpio_pin_handle;
in bbb_gpio.h:
typedef struct {
  gpio_pin_handle super;
  ... specific bbb stuff;
} bbb_pin_handle;
ketul
struct gpio_pin_handle isn't worked
yup that sounds good
@gedare
it means you should create structs of type bbb_pin_handle
and then you pass those to the gpio interface functions
ok now i have to go. mock that together and see what you come up with i suppose. happy hacking!
ketul
yes definitions need to be there in gpio.h otherwise it won't work
okk
bye