Thread: [Cppcms-users] Kindly assist

Brought to you by: artyom-beilis

cppcms-users

[Cppcms-users] Kindly assist

From: M-e-ch T. <mec...@gm...> - 2020-09-04 12:39:47

*[ 29%] Linking C executable external_session_test*
*/data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_sub_4'*
*/data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_load_4'*
*/data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_add_4'*
*clang-10: error: linker command failed with exit code 1 (use -v to see
invocation)*
*make[2]: *** [CMakeFiles/external_session_test.dir/build.make:110:
external_session_test] Error 1*
*make[1]: *** [CMakeFiles/Makefile2:314:
CMakeFiles/external_session_test.dir/all] Error 2*
*make: *** [Makefile:182: all] Error 2*



I am to cppcms and c++ in general how do I solve the above error

On Tue, Sep 1, 2020, 11:17 PM Artyom Beilis <art...@gm...> wrote:

> Hello Dear CppCMS users
>
> Have anybody used raw or asynchronous_raw with http_response?
>
> I'm doing some changes for HTTP/1.1+websockets+performance improvement
> support.
> The original CppCMS design had an underlying assumption that the basic
> communication protocol is derivative of CGI protocols - FastCGI, SCGI.
>
> So when http web server was added it actually parsed CGI output headers
> and converted them to HTTP ones. Thus use of raw/asynchronos_raw mode when
> the user writes his own CGI headers instead of generating them via
> http_response API was reasonable.
>
> Today I want to make an HTTP protocol 1st class citizen in CppCMS - which
> means that I don't silently assume that all IO is CGI based. It is actually
> a requirement for extensions like WebSockets and other things that can't be
> passed via *CGI like protocols.
>
> Removing "raw" *CGI I/O would simplify the task. I can implement it via an
> intermediate CGI parser but it isn't the most trivial task.
>
> So I need to understand if anybody actually uses this feature or I can
> remove it in CppCMS 2.0.0?
>
> Artyom Beilis
> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>

Re: [Cppcms-users] Kindly assist

From: Artyom B. <art...@gm...> - 2020-09-05 21:53:38

Attachments: atomic.patch

Hi,

Thanks for the bug report.

What is the platform you are compiling for x64/arm etc?
What version of clang?

>From googling it looks like atomc library link is missing

Can you try the attached patch and tell if it helps?

Artyom



On Fri, Sep 4, 2020 at 3:40 PM M-e-ch Technologies <
mec...@gm...> wrote:

> *[ 29%] Linking C executable external_session_test*
> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_sub_4'*
> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_load_4'*
> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_add_4'*
> *clang-10: error: linker command failed with exit code 1 (use -v to see
> invocation)*
> *make[2]: *** [CMakeFiles/external_session_test.dir/build.make:110:
> external_session_test] Error 1*
> *make[1]: *** [CMakeFiles/Makefile2:314:
> CMakeFiles/external_session_test.dir/all] Error 2*
> *make: *** [Makefile:182: all] Error 2*
>
>
>
> I am to cppcms and c++ in general how do I solve the above error
>
> On Tue, Sep 1, 2020, 11:17 PM Artyom Beilis <art...@gm...>
> wrote:
>
>> Hello Dear CppCMS users
>>
>> Have anybody used raw or asynchronous_raw with http_response?
>>
>> I'm doing some changes for HTTP/1.1+websockets+performance improvement
>> support.
>> The original CppCMS design had an underlying assumption that the basic
>> communication protocol is derivative of CGI protocols - FastCGI, SCGI.
>>
>> So when http web server was added it actually parsed CGI output headers
>> and converted them to HTTP ones. Thus use of raw/asynchronos_raw mode when
>> the user writes his own CGI headers instead of generating them via
>> http_response API was reasonable.
>>
>> Today I want to make an HTTP protocol 1st class citizen in CppCMS - which
>> means that I don't silently assume that all IO is CGI based. It is actually
>> a requirement for extensions like WebSockets and other things that can't be
>> passed via *CGI like protocols.
>>
>> Removing "raw" *CGI I/O would simplify the task. I can implement it via
>> an intermediate CGI parser but it isn't the most trivial task.
>>
>> So I need to understand if anybody actually uses this feature or I can
>> remove it in CppCMS 2.0.0?
>>
>> Artyom Beilis
>> _______________________________________________
>> Cppcms-users mailing list
>> Cpp...@li...
>> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>>
> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>

Re: [Cppcms-users] Kindly assist

From: M-e-ch T. <mec...@gm...> - 2020-09-06 02:33:33

Attachments: mylog.txt

the patch did better though the same problem reincured at 36%.
Clang and platform are in the log file.
Thank you for taking your time to help me .
Am fairly new to programming


On Fri, Sep 4, 2020, 3:39 PM M-e-ch Technologies <
mec...@gm...> wrote:

> *[ 29%] Linking C executable external_session_test*
> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_sub_4'*
> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_load_4'*
> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_add_4'*
> *clang-10: error: linker command failed with exit code 1 (use -v to see
> invocation)*
> *make[2]: *** [CMakeFiles/external_session_test.dir/build.make:110:
> external_session_test] Error 1*
> *make[1]: *** [CMakeFiles/Makefile2:314:
> CMakeFiles/external_session_test.dir/all] Error 2*
> *make: *** [Makefile:182: all] Error 2*
>
>
>
> I am to cppcms and c++ in general how do I solve the above error
>
> On Tue, Sep 1, 2020, 11:17 PM Artyom Beilis <art...@gm...>
> wrote:
>
>> Hello Dear CppCMS users
>>
>> Have anybody used raw or asynchronous_raw with http_response?
>>
>> I'm doing some changes for HTTP/1.1+websockets+performance improvement
>> support.
>> The original CppCMS design had an underlying assumption that the basic
>> communication protocol is derivative of CGI protocols - FastCGI, SCGI.
>>
>> So when http web server was added it actually parsed CGI output headers
>> and converted them to HTTP ones. Thus use of raw/asynchronos_raw mode when
>> the user writes his own CGI headers instead of generating them via
>> http_response API was reasonable.
>>
>> Today I want to make an HTTP protocol 1st class citizen in CppCMS - which
>> means that I don't silently assume that all IO is CGI based. It is actually
>> a requirement for extensions like WebSockets and other things that can't be
>> passed via *CGI like protocols.
>>
>> Removing "raw" *CGI I/O would simplify the task. I can implement it via
>> an intermediate CGI parser but it isn't the most trivial task.
>>
>> So I need to understand if anybody actually uses this feature or I can
>> remove it in CppCMS 2.0.0?
>>
>> Artyom Beilis
>> _______________________________________________
>> Cppcms-users mailing list
>> Cpp...@li...
>> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>>
>

Re: [Cppcms-users] Kindly assist

From: Artyom B. <art...@gm...> - 2020-09-06 05:26:38

Ok I see,

it looks like -latomic need to be at all levels including user's buid, so I
moved atomic to internal code instead of inline like it was before

I committed some changes to fix the issue can you pease test latest git /
*cpp11* branch

git clone https://github.com/artyom-beilis/cppcms.git
git checkout cpp11

Artyom


On Sun, Sep 6, 2020 at 5:33 AM M-e-ch Technologies <
mec...@gm...> wrote:

> the patch did better though the same problem reincured at 36%.
> Clang and platform are in the log file.
> Thank you for taking your time to help me .
> Am fairly new to programming
>
>
> On Fri, Sep 4, 2020, 3:39 PM M-e-ch Technologies <
> mec...@gm...> wrote:
>
>> *[ 29%] Linking C executable external_session_test*
>> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
>> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_sub_4'*
>> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
>> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_load_4'*
>> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
>> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_add_4'*
>> *clang-10: error: linker command failed with exit code 1 (use -v to see
>> invocation)*
>> *make[2]: *** [CMakeFiles/external_session_test.dir/build.make:110:
>> external_session_test] Error 1*
>> *make[1]: *** [CMakeFiles/Makefile2:314:
>> CMakeFiles/external_session_test.dir/all] Error 2*
>> *make: *** [Makefile:182: all] Error 2*
>>
>>
>>
>> I am to cppcms and c++ in general how do I solve the above error
>>
>> On Tue, Sep 1, 2020, 11:17 PM Artyom Beilis <art...@gm...>
>> wrote:
>>
>>> Hello Dear CppCMS users
>>>
>>> Have anybody used raw or asynchronous_raw with http_response?
>>>
>>> I'm doing some changes for HTTP/1.1+websockets+performance improvement
>>> support.
>>> The original CppCMS design had an underlying assumption that the basic
>>> communication protocol is derivative of CGI protocols - FastCGI, SCGI.
>>>
>>> So when http web server was added it actually parsed CGI output headers
>>> and converted them to HTTP ones. Thus use of raw/asynchronos_raw mode when
>>> the user writes his own CGI headers instead of generating them via
>>> http_response API was reasonable.
>>>
>>> Today I want to make an HTTP protocol 1st class citizen in CppCMS -
>>> which means that I don't silently assume that all IO is CGI based. It is
>>> actually a requirement for extensions like WebSockets and other things that
>>> can't be passed via *CGI like protocols.
>>>
>>> Removing "raw" *CGI I/O would simplify the task. I can implement it via
>>> an intermediate CGI parser but it isn't the most trivial task.
>>>
>>> So I need to understand if anybody actually uses this feature or I can
>>> remove it in CppCMS 2.0.0?
>>>
>>> Artyom Beilis
>>> _______________________________________________
>>> Cppcms-users mailing list
>>> Cpp...@li...
>>> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>>>
>> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>

Re: [Cppcms-users] Kindly assist

From: M-e-ch T. <mec...@gm...> - 2020-09-06 08:03:03

Attachments: mytest.log

Thanks make works fine
Just a few errors with make install ,though I think they are related to my
system. I am not sure though

On Sun, Sep 6, 2020, 8:26 AM Artyom Beilis <art...@gm...> wrote:

> Ok I see,
>
> it looks like -latomic need to be at all levels including user's buid, so
> I moved atomic to internal code instead of inline like it was before
>
> I committed some changes to fix the issue can you pease test latest git /
> *cpp11* branch
>
> git clone https://github.com/artyom-beilis/cppcms.git
> git checkout cpp11
>
> Artyom
>
>
> On Sun, Sep 6, 2020 at 5:33 AM M-e-ch Technologies <
> mec...@gm...> wrote:
>
>> the patch did better though the same problem reincured at 36%.
>> Clang and platform are in the log file.
>> Thank you for taking your time to help me .
>> Am fairly new to programming
>>
>>
>> On Fri, Sep 4, 2020, 3:39 PM M-e-ch Technologies <
>> mec...@gm...> wrote:
>>
>>> *[ 29%] Linking C executable external_session_test*
>>> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
>>> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_sub_4'*
>>> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
>>> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_load_4'*
>>> */data/data/com.termux/files/usr/bin/arm-linux-androideabi-ld:
>>> libcppcms.so.2.0.0.beta1: undefined reference to `__atomic_fetch_add_4'*
>>> *clang-10: error: linker command failed with exit code 1 (use -v to see
>>> invocation)*
>>> *make[2]: *** [CMakeFiles/external_session_test.dir/build.make:110:
>>> external_session_test] Error 1*
>>> *make[1]: *** [CMakeFiles/Makefile2:314:
>>> CMakeFiles/external_session_test.dir/all] Error 2*
>>> *make: *** [Makefile:182: all] Error 2*
>>>
>>>
>>>
>>> I am to cppcms and c++ in general how do I solve the above error
>>>
>>> On Tue, Sep 1, 2020, 11:17 PM Artyom Beilis <art...@gm...>
>>> wrote:
>>>
>>>> Hello Dear CppCMS users
>>>>
>>>> Have anybody used raw or asynchronous_raw with http_response?
>>>>
>>>> I'm doing some changes for HTTP/1.1+websockets+performance improvement
>>>> support.
>>>> The original CppCMS design had an underlying assumption that the basic
>>>> communication protocol is derivative of CGI protocols - FastCGI, SCGI.
>>>>
>>>> So when http web server was added it actually parsed CGI output headers
>>>> and converted them to HTTP ones. Thus use of raw/asynchronos_raw mode when
>>>> the user writes his own CGI headers instead of generating them via
>>>> http_response API was reasonable.
>>>>
>>>> Today I want to make an HTTP protocol 1st class citizen in CppCMS -
>>>> which means that I don't silently assume that all IO is CGI based. It is
>>>> actually a requirement for extensions like WebSockets and other things that
>>>> can't be passed via *CGI like protocols.
>>>>
>>>> Removing "raw" *CGI I/O would simplify the task. I can implement it via
>>>> an intermediate CGI parser but it isn't the most trivial task.
>>>>
>>>> So I need to understand if anybody actually uses this feature or I can
>>>> remove it in CppCMS 2.0.0?
>>>>
>>>> Artyom Beilis
>>>> _______________________________________________
>>>> Cppcms-users mailing list
>>>> Cpp...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>>>>
>>> _______________________________________________
>> Cppcms-users mailing list
>> Cpp...@li...
>> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>>
> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users
>