avr-gcc-list
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [avr-gcc-list] strange c++ obiect initialization problems (full code


From: Georg-Johann Lay
Subject: Re: [avr-gcc-list] strange c++ obiect initialization problems (full code included)
Date: Sun, 12 Jun 2011 18:11:31 +0200
User-agent: Mozilla Thunderbird 1.0.7 (Windows/20050923)

Marcin S schrieb:
2011/6/11 Georg-Johann Lay <address@hidden>:
Marcin S schrieb:
2011/6/9 Georg-Johann Lay <address@hidden>:
Marcin S schrieb:

Hello there,

from time to time i have a strange problems when initializing
 class objects in file range by their respective
constructors, i noted that problem occurs only when that
constructors takes pointers to other global objects, tables
or something else.

Whats the problem? in best scenario object pointer are
wrongly set and pointing to some random locations, at worst
my program does not start at all (event with my main() func
empty) it just hangs

here is link to my code
http://dl.dropbox.com/u/6883398/Lays.tar.bz2 its and eclipse project, simple "make" in project directory should work too.
Note that archive contains already compiled program in
"Release" directory, this compilation does not work at all!
(worst case scenario described above).

As I wrote this program incrementaly it stopped working after
i added 2 objects inits in layouts.cpp

CLayPGMString minTag(minTagData,strMin,true,CLayPGMString::XSMALL); CLayPGMString maxTag(maxTagData,strMax,true,CLayPGMString::XSMALL);

strMin and strMax params are pointers to strings in PROGMEM,
if i replace it with zeros like

CLayPGMString
minTag(minTagData,0,true,CLayPGMString::XSMALL); CLayPGMString
maxTag(maxTagData,0,true,CLayPGMString::XSMALL);

program starts again.

Other important thing noted is, when I initialise those
objects in file range with their default, empty constructors
and then set needed pointers at run time (for example right
after main() function) everything works too. I ran out of
ideas what could be wrong, maybe wrong code, I'm open to any
suggestions.

I'm using avr-gcc-4.5.2 on gentoo linux, also checked on
stable 4.4 and it behaves exactly the same way.

Marcin

Do you see r20 used in one of the functions? Reminds me of http://gcc.gnu.org/PR45263 which is not yet backported to
avr-gcc < 4.6

Johann

Hej Johann,

I'm not sure if i checked it correctly, I've opened .lss file corresponding to my program and yes, there are some r20
operations in this function.

Excuse me my lack of knowledge of low lvl stuff, but You think
this might cause the problem?

That depends on the hardware you use. The problem will not occur on
devices <= 64 kib flash. Look at the code again. In the startup
code that runs before main you might see use of r20 around calls to
you function. Note that these calls will be indirect, so searching
for symbol will not help. Using r20 that way in startup can cause
problems because is is callee-used register and the callee can
clobber it, leaving its content in a mess in startup code.

Marcin

Johann

Hi, It's atmega128, so r20 still is a main suspect.

Note that these calls will be indirect, so searching for symbol
will not help.

Umm, then I'm afraid I dont know how to check it.

Btw. I tried to confirm this thinking by building 4.6.1 avr-gcc but i
think its buggy atm (dwarf2 bug) and i am unable to build any

Yes, that's PR48459: DWARF is unusable in current 4.6 and 4.7, so using DWARF or configuring with --with-dwarf2 will abort.

If you build the compiler on your own, you can simply copy ./gcc/config/avr/libgcc.S from 4.6 or 4.7 into your 4.5 sources. Or you can build 4.6 without dwarf and copy the generated libgcc.a's in respective places in your installed compiler (no matter if win32 or linux).

project with it. I'll wait till it will be usable and let You know.

Marcin.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]