OS hang when multi-process build fail

OS hang when multi-process build fail

[Roumen Petrov]

Hello Cygwin users,

On non regular basis I would like to perform regression test of recent pkix-ssh version with recent openssl release.
In this case is was openssl master branch now release as 3.0.

The problem is that with current cygwin packages gcc make and etc. openssl build with "num of cpu + 1" processes fail with communication error.
Issue is that in such case operation system is unusable.  It respond on mouse events but build process in cygwin console cannot be canceled by keyboard interrupt signal (crtl+c).  Also cygwin console window cannot be closed as well. More over "task manager" and etc cannot be started and I must shutdown OS - [power]  button works.

Builds were run  in cygwin console without administrative rights.

It seems to me current cygwin release has issue with inter-process communication, for instance in setup program as well.


Regards,
Roumen Petrov

P.S. Now I use single process "night"-build as work-around.

Re: OS hang when multi-process build fail

[Marco Atzeri]

On 04.12.2021 09:51, Roumen Petrov wrote:
> Hello Cygwin users,
> 
> On non regular basis I would like to perform regression test of recent 
> pkix-ssh version with recent openssl release.
> In this case is was openssl master branch now release as 3.0.
> 
> The problem is that with current cygwin packages gcc make and etc. 
> openssl build with "num of cpu + 1" processes fail with communication 
> error.
> Issue is that in such case operation system is unusable.  It respond on 
> mouse events but build process in cygwin console cannot be canceled by 
> keyboard interrupt signal (crtl+c).  Also cygwin console window cannot 
> be closed as well. More over "task manager" and etc cannot be started 
> and I must shutdown OS - [power]  button works.

That is strange

> 
> Builds were run  in cygwin console without administrative rights.
> 
> It seems to me current cygwin release has issue with inter-process 
> communication, for instance in setup program as well.


is cygserver running ?

> Regards,
> Roumen Petrov
> 
> P.S. Now I use single process "night"-build as work-around.

Re: OS hang when multi-process build fail

[Roumen Petrov]

Hello,

Please find some details from test today below.
Marco Atzeri wrote:
> On 04.12.2021 09:51, Roumen Petrov wrote:
>> Hello Cygwin users,
>>
>> On non regular basis I would like to perform regression test of recent pkix-ssh version with recent openssl release.
>> In this case is was openssl master branch now release as 3.0.
>>
>> The problem is that with current cygwin packages gcc make and etc. openssl build with "num of cpu + 1" processes fail with communication error.
>> Issue is that in such case operation system is unusable.  It respond on mouse events but build process in cygwin console cannot be canceled by keyboard interrupt signal (crtl+c).  Also cygwin console window cannot be closed as well. More over "task manager" and etc cannot be started and I must shutdown OS - [power]  button works.
>
> That is strange
Yes!

It seems to me network related program stop to work immediately. Existing windows could be switched but no new window.


>>
>> Builds were run  in cygwin console without administrative rights.
>>
>> It seems to me current cygwin release has issue with inter-process communication, for instance in setup program as well.
>
> is cygserver running ?

$ sc  query  | grep -i cyg

Remark: empty list


More details about system:

$ uname -a

CYGWIN_NT-10.0 .... 3.3.2(0.341/5/3) 2021-11-08 16:55 x86_64 Cygwin


$ systeminfo.exe
....
OS Name:                   Microsoft Windows 10 Enterprise
OS Version:                10.0.19042 N/A Build 19042
OS Manufacturer:           Microsoft Corporation
OS Configuration:          Standalone Workstation
OS Build Type:             Multiprocessor Free
....


$ cat /proc/cpuinfo

....
processor       : 7
vendor_id       : AuthenticAMD
cpu family      : 23
model           : 17
model name      : AMD Ryzen 5 PRO 2500U w/ Radeon Vega Mobile Gfx
stepping        : 0
....


$ gcc --version -v
Using built-in specs.
COLLECT_AS_OPTIONS='--version'
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-pc-cygwin/11/lto-wrapper.exe
gcc (GCC) 11.2.0


---- OpenSSL 3.0 bootstrap ----
./Configure Cygwin-x86_64

---- make -j 9 ----
....
gcc  -I. -Iinclude -Iproviders/common/include -Iproviders/implementations/include  -DAES_ASM -DBSAES_ASM -DCMLL_ASM -DECP_NISTZ256_ASM -DGHASH_ASM -DKECCAK1600_ASM -DMD5_ASM -DOPENSSL_BN_ASM_GF2m -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_MONT5 -DOPENSSL_CPUID_OBJ -DOPENSSL_IA32_SSE2 -DPOLY1305_ASM -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DVPAES_ASM -DWHIRLPOOL_ASM -DX25519_ASM  -Wall -O3 -DTERMIOS -DL_ENDIAN -DOPENSSL_PIC -DOPENSSLDIR="\"/usr/local/ssl\"" -DENGINESDIR="\"/usr/local/lib/engines-3\"" -DMODULESDIR="\"/usr/local/lib/ossl-modules\"" -DOPENSSL_BUILDING_OPENSSL -DNDEBUG  -MMD -MF crypto/cms/libcrypto-lib-cms_rsa.d.tmp -MT crypto/cms/libcrypto-lib-cms_rsa.obj -c -o crypto/cms/libcrypto-lib-cms_rsa.obj crypto/cms/cms_rsa.c
       0 [main] make 386 child_info::sync: wait failed, pid 22360, Win32 error 0
    5614 [main] make 386 dofork: child -1 - forked process 22360 died unexpectedly, retry 0, exit code 0x8, errno 11
make[1]: *** [Makefile:14242: crypto/cms/libcrypto-lib-cms_enc.obj] Error 127
make[1]: *** Waiting for unfinished jobs....
---- ctrl+c ----
make[1]: *** Deleting file 'crypto/cms/libcrypto-lib-cms_err.obj'
make[1]: *** [Makefile:14288: crypto/cms/libcrypto-lib-cms_lib.obj] Interrupt
make[1]: *** [Makefile:14272: crypto/cms/libcrypto-lib-cms_io.obj] Interrupt
[no prompt]

Remark: now error is not communication related.


[SNIP]


Regards,
Roumen