NutVersionString on AVR

[Uwe Bonnes]

Hello,

NutVersionString() returns a normal string, needing something like 8 bytes
in RAM space.  On boards woth no external RAM, this is a noticable amount.

What about appended patch?

Bye
--
Uwe Bonnes                bon@...

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
Index: nut/os/version.c
===================================================================
--- nut/os/version.c (Revision 2740)
+++ nut/os/version.c (Arbeitskopie)
@@ -255,8 +255,12 @@
 #include <sys/version.h>
 
 #define OS_VERSION_NUMBER  0x04090700UL
-static CONST char os_version_string[] = "4.9.7.0";
-
+#ifdef PROGMEM
+static CONST char os_version_string[] PROGMEM
+#else
+static CONST char os_version_string[]
+#endif
+ = "4.9.7.0";
 /*!
  * \addtogroup xgNutVersion
  */
_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Uwe Bonnes]

Hello,

some days ago I proposed to put  os_version_string into progmem, when
progmem is available, like
+#ifdef PROGMEM
+static CONST char os_version_string[] PROGMEM
+#else
+static CONST char os_version_string[]
+#endif
+ = "4.9.7.0";
 
Would this be transparent for a usage like
./nut/pro/httpd.c:    fprintf_P(stream, fmt_P, HTTP_MAJOR_VERSION, HTTP_MINOR_VERSION, status, title, NutVersionString());

Bye
--
Uwe Bonnes                bon@...

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Ethernut]

Hi Uwe,

Uwe Bonnes wrote:

> +#ifdef PROGMEM
> +static CONST char os_version_string[] PROGMEM
> +#else
> +static CONST char os_version_string[]
> +#endif
> + = "4.9.7.0";

The #ifs are not required, because non-Harvard architectures provide a
PROGMEM dummy.


> Would this be transparent for a usage like
> ./nut/pro/httpd.c:    fprintf_P(stream, fmt_P, HTTP_MAJOR_VERSION, HTTP_MINOR_VERSION, status, title, NutVersionString());

It would be incompatible, because %P is required instead of %s. This
would break a large number of applications, just to save 8 bytes of RAM.

Harald

_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Uwe Bonnes]

>>>>> "Harald" == Harald Kipp <harald.kipp@...> writes:

    Harald> Hi Uwe, Uwe Bonnes wrote:

    >> +#ifdef PROGMEM +static CONST char os_version_string[] PROGMEM +#else
    >> +static CONST char os_version_string[] +#endif + = "4.9.7.0";

    Harald> The #ifs are not required, because non-Harvard architectures
    Harald> provide a PROGMEM dummy.


    >> Would this be transparent for a usage like ./nut/pro/httpd.c:
    >> fprintf_P(stream, fmt_P, HTTP_MAJOR_VERSION, HTTP_MINOR_VERSION,
    >> status, title, NutVersionString());

    Harald> It would be incompatible, because %P is required instead of
    Harald> %s. This would break a large number of applications, just to
    Harald> save 8 bytes of RAM.

What about adding NutVersionString_P() then?

--
Uwe Bonnes                bon@...

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Uwe Bonnes]

>>>>> "Harald" == Harald Kipp <harald.kipp@...> writes:

    Harald> Hi Uwe, Uwe Bonnes wrote:

    >> +#ifdef PROGMEM +static CONST char os_version_string[] PROGMEM +#else
    >> +static CONST char os_version_string[] +#endif + = "4.9.7.0";

    Harald> The #ifs are not required, because non-Harvard architectures
    Harald> provide a PROGMEM dummy.


    >> Would this be transparent for a usage like ./nut/pro/httpd.c:
    >> fprintf_P(stream, fmt_P, HTTP_MAJOR_VERSION, HTTP_MINOR_VERSION,
    >> status, title, NutVersionString());

    Harald> It would be incompatible, because %P is required instead of
    Harald> %s. This would break a large number of applications, just to
    Harald> save 8 bytes of RAM.

Do non-Harvard architectures provide a dummy %P?
--
Uwe Bonnes                bon@...

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Ethernut]

Uwe Bonnes wrote:

> Do non-Harvard architectures provide a dummy %P?

Yes, it handles %P like %s.

Harald

_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Ethernut]

Uwe Bonnes wrote:
> What about adding NutVersionString_P() then?

How would that help? Compatibility requires to have NutVersionString
returning a string in RAM. Adding NutVersionString_P would add more
code, nothing else.

May I ask why you spend that much time on 8 bytes of RAM? I'm almost
sure there are more promising parts in the library.

Harald

_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion

[Uwe Bonnes]

>>>>> "Harald" == Harald Kipp <harald.kipp@...> writes:

    Harald> Uwe Bonnes wrote:
    >> What about adding NutVersionString_P() then?

    Harald> How would that help? Compatibility requires to have
    Harald> NutVersionString returning a string in RAM. Adding
    Harald> NutVersionString_P would add more code, nothing else.

    Harald> May I ask why you spend that much time on 8 bytes of RAM? I'm
    Harald> almost sure there are more promising parts in the library.

Well on the AT90CAN128 with no external RAM, code memory is ample and RAM is
few. Adding a printf_P with the version string brought down available heap,
so I was thinking about how to keep that usefill feature witout scrificing
RAM...
But thinking again, probably most usefull would be to have
#define OS_VERSION_NUMBER  0x04090700UL
in sys/version.h an so one could construct PROGMEM strings with it and use
it for conditional compilations.
Sorry for wasting your time.

Bye

--
Uwe Bonnes                bon@...

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
_______________________________________________
http://lists.egnite.de/mailman/listinfo/en-nut-discussion