www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Got some problemmes with new version of compiler 2.090

reply uranuz <neuranuz gmail.com> writes:
Hello! I have got some problemmes with new compiler version that 
I am trying to debug already for several days, because I cannot 
locate the source of bug using GDB, because there are no useful 
backtrace.
The first problemme was `Memory allocation failed`.
I was executing my code step by step. And it was failing with 
such error near where I was using std.exception: enforce.
I need to say that I running server in threaded mode using 
TaskPool class. But when running in `plain` mode without threads 
all goes just fine.
What changes are made in enforce, TaskPool or memory allocation 
that can be a source of such bug?

And another error that I have encountered just recently:
//---
Signal        Stop	Print	Pass to program	Description
SIGUSR1       No	No	Yes		User defined signal 1
Signal        Stop	Print	Pass to program	Description
SIGUSR2       No	No	Yes		User defined signal 2
Source directories searched: 
/usr/include/dmd/druntime/import:/home/uranuz/sources/yar_mkk:$cdir:$cwd
Source directories searched: 
/usr/include/dmd/phobos:/usr/include/dmd/druntime/import:/home/uranuz/sources/yar_mkk:$cdir:$cwd
Source directories searched: 
/home/uranuz/sources/webtank/src:/usr/include/dmd/phobos:/usr/include/dmd/druntime/import:/home/uranuz/sources/yar_mkk:$cdir:$cwd
Source directories searched: 
/home/uranuz/sources/ivy/src:/home/uranuz/sources/webtank/src:/usr/include/dmd/phobos:/usr/include/dmd/druntime/import:/home/uranuz/sources/yar_mkk:$cdir:$cwd
Source directories searched: 
/home/uranuz/sources/yar_mkk/src:/home/uranuz/sources/ivy/src:/home/uranuz/sources/webtank/src:/usr/include/dmd/phobos:/usr/include/dmd/druntime/import:/home/uranuz/sources/yar_mkk:$cdir:$cwd
Running executable
[Thread debugging using libthread_db enabled]
Using host libthread_db library 
"/lib/x86_64-linux-gnu/libthread_db.so.1".
[New Thread 0x7ffff7ff6700 (LWP 14697)]
[New Thread 0x7ffff7ff1700 (LWP 14698)]
[New Thread 0x7ffff7fec700 (LWP 14699)]
[New Thread 0x7ffff7fe7700 (LWP 14700)]
[New Thread 0x7ffff7fe2700 (LWP 14701)]
[New Thread 0x7ffff7fdd700 (LWP 14702)]
[New Thread 0x7ffff7fd8700 (LWP 14703)]
[New Thread 0x7ffff7fbc700 (LWP 14704)]
[New Thread 0x7ffff7fb7700 (LWP 14705)]
[New Thread 0x7ffff7fb2700 (LWP 14706)]
[New Thread 0x7ffff7fad700 (LWP 14707)]
[New Thread 0x7ffff7fa8700 (LWP 14708)]
[New Thread 0x7ffff7fa3700 (LWP 14709)]
[New Thread 0x7ffff7f9e700 (LWP 14710)]
[New Thread 0x7ffff7f99700 (LWP 14711)]
warning: Error removing breakpoint 1
warning: Error removing breakpoint 2
warning: Error removing breakpoint 3
warning: Error removing breakpoint 4
warning: Error removing breakpoint 5
warning: Error removing breakpoint 6
warning: Error removing breakpoint 7
[New Thread 0x7ffff177a700 (LWP 14790)]

Thread
1 "mkk_main_servic" received signal SIGTRAP, Trace/breakpoint 
trap.
0x0000555555ce9cd8 in 
webtank.net.server.thread_pool.ThreadPoolServer._runLoop() 
(this=0x7ffff7e9c660) at 
../webtank/src/webtank/net/server/thread_pool.d:61
61					if( client is null )
[Switching to thread 17 (Thread 0x7ffff177a700 (LWP 
14790))](running)

Thread
17 "mkk_main_servic" received signal SIGTRAP, Trace/breakpoint 
trap.
[Switching to Thread 0x7ffff177a700 (LWP 14790)]
0x0000555555d3b79b in 
_D7webtank3net6server6common14processRequestFC3std6socket6SocketCQClQCgQCf
iface10IWebServerZv (server=0x7ffff7e9c670, sock=0x7ffff7f044c0) at
../webtank/src/webtank/net/server/common.d:95
95		HTTPOutput response = new HTTPOutput();

Thread
17 "mkk_main_servic" received signal SIGSEGV, Segmentation fault.
0x0000555555deda1e in _d_newclass ()
bt


_D7webtank3net6server6common14processRequestFC3std6socket6SocketCQClQCgQCf
iface10IWebServerZv (server=0x7ffff7e9c670, sock=0x7ffff7f044c0) at
../webtank/src/webtank/net/server/common.d:95

_D3std11parallelism__T3runTPFCQBc6socket6SocketC7webtank3net6server5iface10IWebServerZvTQChTCQBtQBoQBn11thread_pool16ThreadPoolServ
rZQEiFQEhKQEjKQCcZv (_param_2= 0x7ffff7f005e8: 0x7ffff7e9c660,
_param_1= 0x7ffff7f005e0: 0x7ffff7f044c0, fpOrDelegate=0x555555d3b780
<_D7webtank3net6server6common14processRequestFC3std6socket6SocketCQClQCgQCf5i
ace10IWebServerZv>) at /usr/include/dmd/phobos/std/parallelism.d:761

_D3std11parallelism__T4TaskSQBaQz3runTPFCQBn6socket6SocketC7webtank3net6server5iface10IWebServerZvTQChTCQBtQBoQBn11thread_pool16ThreadPool
erverZQEt4implFPvZv (myTask=0x7ffff7f005a0) at
/usr/include/dmd/phobos/std/parallelism.d:444


_D3std11parallelism8TaskPool5doJobMFPSQBkQBj12AbstractTaskZv ()

std.parallelism.TaskPool.executeWorkLoop() ()

std.parallelism.TaskPool.startWorkLoop() ()



pthread_create.c:463

../sysdeps/unix/sysv/linux/x86_64/clone.S:95
//-----

Seems that some error occured when allocating new class instance. 
It is very weird
Jan 19 2020
next sibling parent uranuz <neuranuz gmail.com> writes:
And also when all `enfroce`ments pass and condition is true then 
there are no `Memory allocation failed` errors. So it the most 
certaily a bug in enforce or with the way it allocates memory for 
exceptions, maybe.
Jan 19 2020
prev sibling next sibling parent reply berni44 <dlang d-ecke.de> writes:
On Sunday, 19 January 2020 at 08:54:43 UTC, uranuz wrote:

 The first problemme was `Memory allocation failed`.


Maybe it could be helpful if you could provide a minimum example, 
e.g. using dustmite [1] to create one.

[1] https://github.com/CyberShadow/DustMite/wiki
Jan 19 2020
parent reply uranuz <neuranuz gmail.com> writes:
Seems that problemme with memory allocation repeats even using 
regular `throw new Exception`. But I don't understand why it is 
only occur in threaded code.
I don't know how to use dustmite unfortunately. Seems that I need 
to formulate some condition to check stdout or stderror. But my 
programme doesn't output anything there. Even if error occurs it 
is being sent by TCPSocket
Jan 19 2020
parent user1234 <user12324 1234.de> writes:
On Sunday, 19 January 2020 at 09:57:09 UTC, uranuz wrote:

 Seems that problemme with memory allocation repeats even using 
 regular `throw new Exception`. But I don't understand why it is 
 only occur in threaded code.
 I don't know how to use dustmite unfortunately.


You can also use https://github.com/CyberShadow/Digger on the non 
minimized code.
The goal is to point to a specific compiler change from which 
your server start failing.
If this works (*) this will already give a better overview of the 
issue, although we already know it's certainly related to 
druntime and more especially the GC.


 Seems that I need to formulate some condition to check stdout 
 or stderror. But my programme doesn't output anything there. 
 Even if error occurs it is being sent by TCPSocket


(*): often it works but points to merge commits from stable 
master, which is not so helpful.
Jan 19 2020
prev sibling parent reply Mathias Lang <pro.mathias.lang gmail.com> writes:
I think this would be better suited for `D.learn` or 
stackoverflow. But the thread is here, so next time :)

On Sunday, 19 January 2020 at 08:54:43 UTC, uranuz wrote:

 The first problemme was `Memory allocation failed`.
 I was executing my code step by step. And it was failing with 
 such error near where I was using std.exception: enforce.
 I need to say that I running server in threaded mode using 
 TaskPool class. But when running in `plain` mode without 
 threads all goes just fine.
 What changes are made in enforce, TaskPool or memory allocation 
 that can be a source of such bug?


This kind of issue is really platform specific, so it would be 
helpful to know which OS you are running, what's the env like 
(How much memory ? Is it virtualized ?), etc...


 And another error that I have encountered just recently:
 [...]


That looks related. Can you start your program with 
`--DRT-gcopts=parallel:0` ?
Alternatively you can use this in your code in one of the module 
that is compiled in:
```
static if (__VERSION__ >= 2087)
     extern(C) __gshared string[] rt_options = [ 
"gcopt=parallel:0" ];
```
Jan 19 2020
next sibling parent uranuz <neuranuz gmail.com> writes:
 This kind of issue is really platform specific, so it would be 
 helpful to know which OS you are running, what's the env like 
 (How much memory ? Is it virtualized ?), etc...


I am using Ubuntu 18.04 with latest updates. It is not 
virtualized (AMD Ryzen 3700 processor). I have 32GB of memory, so 
I am very doubt that there could be any memory lack issue.
Nothing special or specific about it, I believe, because Ubuntu 
is enough popular and wide-spread distributive of Linux. And one 
of the distributives listed on dlang "Downloads" page. So I 
believe that DMD should be tested on this OS.

When I find enough time I shall try manually reduce problemme, 
because I don't have any idea how to apply Dustmite, etc. to this.
But for now seems that I need to revert to older compiler 
version, because I have already spend so much time for things 
that are not related to main goals of my project.

I was just interested if someone of developers remembers what 
changes were made to compiler/ library that may cause this 
problemme?

Thanks
Jan 19 2020
prev sibling next sibling parent uranuz <neuranuz gmail.com> writes:
Seems that problemme was introduced between 2.089.1 and 2.090 
exactly.
Jan 19 2020
prev sibling parent uranuz <neuranuz gmail.com> writes:
I have rewritten bunch of code and now it doesn't crash anymore. 
But still I failed to figure out why it was failing...
Jan 27 2020