does not compile #17

mikrocoder · 2022-02-28T11:19:42Z

Hi,

your lib looks very promising. Unfortunately there are big problems with it. If I use it with Arduino then the controller always resets.

If I use it in Atmel Studio 7 it always compiles with error messages. Your vector example does not compile either. It doesn't matter if I use an avr-gcc toolchain with gcc 9.3.0 or 11.2.0. Also in the Arduino IDE.

I can include vector without errors, but when I use it, it doesn't compile anymore.

Now I wanted to ask in which IDE do you program? In which environment did you test all this? According to your description it should work with avr-gcc.

2 Examples with Atmel Studio 7.

#include <vector>
#include <cstdio>
#include <algorithm>

int main()
{
    std::vector<uint8_t> test{10, 1, 2, 42, 3};
    test.push_back(4);
    test.erase(test.begin());

    if(auto it = std::find(test.begin(), test.end(), 42); it != test.end()) {
        test.erase(it);
    }
}

Output

Error undefined reference to `operator new(unsigned int)'	avr-libstdcpp-Test_001	<artificial>	1
Error <artificial>:(.text.startup+0x3e): undefined reference to `operator new(unsigned int)'	avr-libstdcpp-Test_001		1
Error <artificial>:(.text.startup+0x6c): undefined reference to `operator delete(void*, unsigned int)'	avr-libstdcpp-Test_001		1
Error <artificial>:(.text.startup+0xd2): undefined reference to `operator delete(void*, unsigned int)'	avr-libstdcpp-Test_001		1
Error C:\Users\Worker\AppData\Local\Temp\cc6G2oHd.ltrans0.ltrans.o: in function `main':	avr-libstdcpp-Test_001

#include <avr/io.h>
#include <vector>
#include <algorithm>

using namespace std;

vector<int16_t> v1 {1, 2, 3, 4};
vector<int16_t> v3 (19);

int main()
{  }

Output

Error undefined reference to `operator new(unsigned int)'	avr-libstdcpp-Test_001	<artificial>	1
Error <artificial>:(.text.startup+0x8a): undefined reference to `operator new(unsigned int)'	avr-libstdcpp-Test_001		1
Error C:\Users\Worker\AppData\Local\Temp\ccYoUbRD.ltrans0.ltrans.o: in function `std::_Vector_base<int, std::allocator<int> >::~_Vector_base()':	avr-libstdcpp-Test_001		1
Error C:\Users\Worker\AppData\Local\Temp\ccYoUbRD.ltrans0.ltrans.o: in function `_GLOBAL__sub_I_v1':	avr-libstdcpp-Test_001		1
Error ld returned 1 exit status	avr-libstdcpp-Test_001	collect2.exe	0
Error undefined reference to `operator delete(void*, unsigned int)'	avr-libstdcpp-Test_001	<artificial>	1

Translated with www.DeepL.com/Translator (free version)

The text was updated successfully, but these errors were encountered:

rleh · 2022-02-28T12:14:01Z

Have taken a look at the vector example in the examples/ directory?
You can't just copy the vector header file to you project, include it and expect it to work. There is a Makefile that specifies all compiler flags, include directories and source directories necessary for the avr-libstdc++ to work.

All examples get tested by our CI (Github Action) on every commit automatically and work fine: https://github.com/modm-io/avr-libstdcpp/runs/5250992849?check_suite_focus=true

mikrocoder · 2022-02-28T13:34:32Z

Hi,

I have now specified all the flags according to the make file in Atmel Studio. It still does not compile with the same errors.

About your objection. How else are you supposed to do it? That is the usual way. Make lib folder known, then include and apply.

You tested this only with the g++? But you write it would also work with the avr-gcc. Which it doesn't. Did you also test it with avr-gcc?

rleh · 2022-02-28T14:44:05Z

I have now specified all the flags according to the make file in Atmel Studio. It still does not compile with the same errors.

Then your build probably still misses the source files in $(LIB_DIR)/src/*.cc, see line 18 of the Makefile.

You tested this only with the g++? But you write it would also work with the avr-gcc. Which it doesn't. Did you also test it with avr-gcc?

[avr-]gcc/[avr-]g++ are the exact same compiler, except that a call to g++/avr-g++ enables C++ language support by default. You might have to pass -x c++ to avr-gcc command to switch on C++ language support.

Which exact compiler version are you using? Which environment (operation system, ...) are you using? Where did you get your avr-gcc compiler?
~~Have you tried our precompiled avr-gcc v11.2.0?~~ (Only for Linux, looks like you are using Windows.)

What additional compiler flags does you IDE(s) pass to the avr-gcc compiler?
Have you tried compiling the unmodified vector example from this Repo using the Makefile (by just calling make in the example directory from shell)?

rleh · 2022-02-28T14:50:26Z

Looking at the error messages you posted in you first comment the problem seems easy: You are missing the operator new() and operator delete() implementations.
In the vector example of this repository they come from examples/common/new.cpp.
This file is added to the sources in line 18 of the Makefile ($(COMMON_DIR)/*.cpp), so you probably only have to copy that file (and maybe also cxxabi.cpp and uart.cpp) to you project, ensure the compiler compiles and links them for your error to disappear.

mikrocoder · 2022-03-01T15:06:05Z

Hi,

can we speak in german?

CombiesGit · 2022-03-02T02:00:01Z

I would like to thank you very much for the question.

And thanks for the replies.

So I managed to put this std++ implementation into operation in Arduino.

A few tests still need to be done.
Then I would be happy to provide a short guide

ckormanyos · 2022-04-10T12:32:48Z

Looking at the error messages you posted in you first comment the problem seems easy: You are missing the operator new() and operator delete() implementations.

I just looked at this post and it's going in the right direction. Raphael (@rleh) is correct and on the right track for this issue.

Indeed, one of the tricky things about std::vector is that is uses dynamic menory (such as an allocator that reduces to using operators new/delete under the hood). So

If you look at the full template signature of std::vector, you will see that the second parameter is the allocator used for memory allocation.

Things you can try...

Just to get an example running, you might try the advice from Raphael (@rleh) above.
Sometimes it can be helpful to make a similar example with std::array.
Write/use a custom allocator (this is a slightly advanced topic). This can be a great way to go on the metal.

I have examples for most of this stuff if you'd like to dive deeper into any of these ideas...

mikrocoder · 2022-04-13T16:16:42Z

Hello,

it was about how to get the avr-libstdcpp lib running in Atmel Studio on Windows. I was able to solve this with the help of another forum. If you are interested in this ...

You have to specify the path to the include directory in the project properties under directories.

Under Miscellaneous the usual: -std=c++20 -flto -Wno-volatile
Under Symbol I enter F_CPU.

And in Solution Explorer you have to add the .cc and .cpp files.

After that you can start with the avr-libstdcpp lib.

If you want to show examples, feel free to do so. :-)

ckormanyos · 2022-04-15T10:42:30Z

it was about how to get the avr-libstdcpp lib running in Atmel Studio on Windows

Understood. Sorry, I had not understood that point. Yes. Simply add the include path on the compiler command line with the GUI (which boils down to the GCC command line -I include path syntax).

Not every compiler allows the specification of a path for its STL. GCC, however, odes a fine job on this for AVR and also for other toolchain targets such as ARM. So if you ever get a cool embedable STL, you can use it not only with AVR in Microchip's ATMEL Studio, but also with other targets on other environments.

I do have a related question: I notice you are mentioning the flag -std=c++20. Have you upgraded the GCC tollchain in your ATMEL Studio? Or does a more modern version or ATMEL Studio install a toolchain beyond GCC 5.4 AVR?

ckormanyos · 2022-04-15T11:42:21Z

If you want to show examples, feel free to do so

Hi @mikrocoder in another repository, I have some detailed benchmarks running on AVR. Some of them make rather heavy use of C++/STL.

For each benchmark, portd.3 measures the runtime on the scope.

You might enjoy the depth and scope of some of those benchmarks and their enclosing repo. In the root of theproject, there is a workspace for ATMEL Studio.

mikrocoder · 2022-04-15T13:42:07Z

Hi,

I'll have a look at your examples. Let's see if I understand them. Anyway, thanks for that.

AS always uses its outdated avr-gcc 5.4.0. There is nothing newer. I use my own toolchain with avr-gcc 11.2.0 with binutils 2.38. From the http://packs.download.atmel.com/ are added, see below ... the standard controllers are included anyway.

In Atmel Studio you still have to add the packages in Device Manager. If you want can upload my toolchain to dropbox. Can be included in Atmel Studio and also in Arduino IDE.

Zak Kemble https://blog.zakkemble.net/avr-gcc-builds/ also offers updated toolchains in larger intervals. Only his description for Arduino does not work.

atmega328pb

atmega808 	
atmega809 	
atmega1608 		
atmega1609
atmega3208 		
atmega3209 	
atmega4808 	
atmega4809 	

attiny202 	
attiny204 	
attiny212 	
attiny214 	
attiny402 	
attiny404 	
attiny406 	
attiny412 	
attiny414 	
attiny416 	
attiny417 	
attiny804 	
attiny806 	
attiny807 	
attiny814 	
attiny816 	
attiny817 	
attiny1604 			
attiny1606 			
attiny1607 			
attiny1614
attiny1616
attiny1617
attiny3216
attiny3217
attiny1624
attiny1626
attiny1627
attiny3224
attiny3226
attiny3227

avr32da28
avr32da32
avr32da48
avr64da28
avr64da32
avr64da48
avr64da64
avr128da28
avr128da32
avr128da48
avr128da64

avr32db28
avr32db32
avr32db48
avr64db28
avr64db32
avr64db48
avr64db64
avr128db28
avr128db32
avr128db48
avr128db64

avr16dd14
avr16dd20
avr16dd28
avr16dd32
avr32dd14
avr32dd20
avr32dd28
avr32dd32
avr64dd14
avr64dd20
avr64dd28
avr64dd32

Translated with www.DeepL.com/Translator (free version)

ckormanyos · 2022-04-15T13:50:03Z

I use my own toolchain with avr-gcc 11.2.0 with binutils 2.38

That is a really good idea. I am glad to hear oyu did that. I'll get back to you if needed.

We are straying just a bit off topic, but for more completeness, one more link...

You might like my build of avr-gcc found here as a multi-file ZIP archive for windows (but you'll need a copy of libwinpthread-1.dll).

I solved some of the missing device file problems (but not all of them). I made a bunch of notes on building here.

There are some other target GCC builds there that might interest you as well.

mikrocoder · 2022-05-06T17:22:46Z

Hello,

there is a new problem. gcc 11.3.0 is out. In it there were changes in the math lib. Now sadly with avr-libstdcpp there is the problem of conflictings declarations and previous declarations. Could this be fixed at some point?
My test environment is Windows 10 64bit, Microchip Studio 7 with my toolchains. With avr-gcc 11.2.0 everything was still ok.

Severity	Code	Description	Project	File	Line
Message		'float copysignf(float, float)' previously defined here	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	389
Message		'float fabsf(float)' previously defined here	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	163
Error		conflicting declaration of C function 'bool isfinitef(float)'	avrLibStdCpp	C:\avrToolchain\avrLibStdCpp\include\cmath	144
Error		conflicting declaration of C function 'bool isfinitef(float)'	avrLibStdCpp	C:\Users\Worker\Documents\Atmel Studio\7.0\WorkSpace_AVR128DB48\avrLibStdCpp\avrLibStdCpp\math.cc	142
Error		conflicting declaration of C function 'bool isinff(float)'	avrLibStdCpp	C:\avrToolchain\avrLibStdCpp\include\cmath	143
Error		conflicting declaration of C function 'bool isinff(float)'	avrLibStdCpp	C:\Users\Worker\Documents\Atmel Studio\7.0\WorkSpace_AVR128DB48\avrLibStdCpp\avrLibStdCpp\math.cc	137
Error		conflicting declaration of C function 'bool isnanf(float)'	avrLibStdCpp	C:\avrToolchain\avrLibStdCpp\include\cmath	142
Error		conflicting declaration of C function 'bool isnanf(float)'	avrLibStdCpp	C:\Users\Worker\Documents\Atmel Studio\7.0\WorkSpace_AVR128DB48\avrLibStdCpp\avrLibStdCpp\math.cc	132
Message		previous declaration 'int isfinitef(float)'	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	365
Message		previous declaration 'int isfinitef(float)'	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	365
Message		previous declaration 'int isinff(float)'	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	358
Message		previous declaration 'int isinff(float)'	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	358
Message		previous declaration 'int isnanf(float)'	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	348
Message		previous declaration 'int isnanf(float)'	avrLibStdCpp	c:\avrtoolchain\avr-gcc-11.3.0_mingw32_binutils2.38\avr\include\math.h	348
Error		recipe for target 'math.o' failed	avrLibStdCpp	C:\Users\Worker\Documents\Atmel Studio\7.0\WorkSpace_AVR128DB48\avrLibStdCpp\avrLibStdCpp\Release\Makefile	141
Error		redefinition of 'float copysignf(float, float)'	avrLibStdCpp	C:\Users\Worker\Documents\Atmel Studio\7.0\WorkSpace_AVR128DB48\avrLibStdCpp\avrLibStdCpp\math.cc	147
Error		redefinition of 'float fabsf(float)'	avrLibStdCpp	C:\Users\Worker\Documents\Atmel Studio\7.0\WorkSpace_AVR128DB48\avrLibStdCpp\avrLibStdCpp\math.cc	32

ckormanyos · 2022-05-06T19:53:58Z

gcc 11.3.0 is out

Do you have a build of avr-gcc 11.3 that you're using?

Now sadly with avr-libstdcpp there is the problem of conflictings declarations and previous declarations. Could this be fixed at some point?

This topic is something i can look into. But it might take a bit of time to straighten out the float, double and long double C/C++ functions for both 32 and 64 bit. It is something I've looked into before, but not fully completed yet.

mikrocoder · 2022-05-06T20:08:21Z

Hi,

gcc 11.3.0 is out

Do you have a build of avr-gcc 11.3 that you're using?

Yes.

Now sadly with avr-libstdcpp there is the problem of conflictings declarations and previous declarations. Could this be fixed at some point?

This topic is something i can look into. But it might take a bit of time to straighten out the float, double and long double C/C++ functions for both 32 and 64 bit. It is something I've looked into before, but not fully completed yet.

Okay that would be very nice. It does not push. I have plenty of time.

ckormanyos · 2022-05-07T17:55:23Z

Do you have a build of avr-gcc 11.3 that you're using?

Yes.

Cool. I actually mean, do you have one available that we could use for testing? I can spin one up in a few hours, but if you've got one, that's easier...

mikrocoder · 2022-05-07T21:15:04Z

Hi,

I can provide. For Windows in .rar and .zip, each as he likes and for Linux. Whereby the Linux version is not tested. I use only the Windows version. I delete again in a few days. Link

ckormanyos · 2022-05-08T10:08:33Z

I can provide. For Windows in .rar and .zip, each as he likes and for Linux.

Awesome builds @mikrocoder. Many thanks.

OK, so here is the deal... Lines like this are actually incorrect. This is because the proper C-language signature of such functions returns int instead of bool.

I have not figured out all the inconsistencies yet. And I'm not sure if compiler option or complete repair is best (as these signatures originate in <math.h>.

So I have opened a new issue #18 to handle this. I'll try a few things there and report back when done.

ckormanyos · 2022-05-08T10:12:40Z

Can we now close this issue @mikrocoder (assuming we follow the latest point in #18) ?

Cc: @salkinium and @rleh and @chris-durand

salkinium · 2022-05-08T11:08:03Z

I've given you triage rights @ckormanyos, so that you can manage issues directly.

ckormanyos · 2022-05-08T11:42:12Z

The first point is finished:

When using this STL library with ATMEL Studio, two thngs are needed:

Use ATMEL Studio with a compiler version higher than the installed version (such as get above GCC 8).
Add (via -I compiler command line option) the include path of avr-libstdcpp to the command line.

The second point (<cmath> in cimbination with a new GCC 11.3) is being handled in #18.

ckormanyos · 2022-05-08T12:55:09Z

given you triage rights

Thank you Niklas (@salkinium).

mikrocoder · 2023-10-12T09:27:31Z

I think I forgot to thank you for your work. Better late than never. :-)

ckormanyos mentioned this issue May 8, 2022

Repair problems ISO C signatures in <cmath> #18

Closed

ckormanyos closed this as completed May 8, 2022

ckormanyos mentioned this issue May 9, 2022

Handle ISO C <cmath> fixes and add example #20

Merged

mikrocoder mentioned this issue Oct 12, 2023

Question about changes to core files (Information Request) SpenceKonde/DxCore#481

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

does not compile #17

does not compile #17

mikrocoder commented Feb 28, 2022

rleh commented Feb 28, 2022 •

edited

Loading

mikrocoder commented Feb 28, 2022

rleh commented Feb 28, 2022

rleh commented Feb 28, 2022

mikrocoder commented Mar 1, 2022

CombiesGit commented Mar 2, 2022

ckormanyos commented Apr 10, 2022

mikrocoder commented Apr 13, 2022

ckormanyos commented Apr 15, 2022

ckormanyos commented Apr 15, 2022

mikrocoder commented Apr 15, 2022

ckormanyos commented Apr 15, 2022 •

edited

Loading

mikrocoder commented May 6, 2022 •

edited

Loading

ckormanyos commented May 6, 2022

mikrocoder commented May 6, 2022 •

edited

Loading

ckormanyos commented May 7, 2022

mikrocoder commented May 7, 2022

ckormanyos commented May 8, 2022 •

edited

Loading

ckormanyos commented May 8, 2022

salkinium commented May 8, 2022

ckormanyos commented May 8, 2022

ckormanyos commented May 8, 2022

mikrocoder commented Oct 12, 2023

does not compile #17

does not compile #17

Comments

mikrocoder commented Feb 28, 2022

rleh commented Feb 28, 2022 • edited Loading

mikrocoder commented Feb 28, 2022

rleh commented Feb 28, 2022

rleh commented Feb 28, 2022

mikrocoder commented Mar 1, 2022

CombiesGit commented Mar 2, 2022

ckormanyos commented Apr 10, 2022

mikrocoder commented Apr 13, 2022

ckormanyos commented Apr 15, 2022

ckormanyos commented Apr 15, 2022

mikrocoder commented Apr 15, 2022

ckormanyos commented Apr 15, 2022 • edited Loading

mikrocoder commented May 6, 2022 • edited Loading

ckormanyos commented May 6, 2022

mikrocoder commented May 6, 2022 • edited Loading

ckormanyos commented May 7, 2022

mikrocoder commented May 7, 2022

ckormanyos commented May 8, 2022 • edited Loading

ckormanyos commented May 8, 2022

salkinium commented May 8, 2022

ckormanyos commented May 8, 2022

ckormanyos commented May 8, 2022

mikrocoder commented Oct 12, 2023

rleh commented Feb 28, 2022 •

edited

Loading

ckormanyos commented Apr 15, 2022 •

edited

Loading

mikrocoder commented May 6, 2022 •

edited

Loading

mikrocoder commented May 6, 2022 •

edited

Loading

ckormanyos commented May 8, 2022 •

edited

Loading