[openssl-users] VS2017 15.5 hangs indefinitely with Blake2b.c
googleersatz at oliverniebuhr.de
Wed Dec 6 13:43:12 UTC 2017
Like mentioned below, I updated the VS 2017 Compiler to 14.12 - but it
also happens with 14.11.
Problem is: There is No Error Message of any kind - CL.exe keeps running
(well more or less) indefinitely. It simply stops working.
It really seems like a Compiler Bug. But I am too OpenSSL noobish to be
able to tell what the Problem is from my end.
I will create a BR on Github.
On 06/12/2017 13:06, Matt Caswell wrote:
> On 06/12/17 08:24, Oliver Niebuhr wrote:
>> I am not sure if it is a VS Compiler Bug or a Bug in OpenSSL.
>> When the Compiler reaches 'Blake2b.c', it creates the 'Blake2b.obj' and
>> then hangs indefinitely.
>> CPU usage for CL stays at 27-28 Percent. Used RAM stays at 34-35 Megabyte.
>> OpenSSL Version: 1.1.0 Git up2date, Static and Dynamic build
>> Windows: Windows 10, X64, FCU, up2date
>> Windows 10 SDK: 16299
>> VS Compiler Version: 14.11 and 14.12, running from VS native CMD Line
>> NASM: 2.13.02 (also happens without NASM)
>> Perl Version: Strawberry Perl 126.96.36.199
>> Perl Configure:
>> no-comp no-rc5 no-idea no-mdc2 no-tls1 no-tls1-method no-tls1_1
>> no-tls1_1-method no-dtls no-dtls1-method no-dtls1_2 no-dtls1_2-method
>> no-deprecated no-weak-ssl-ciphers VC-WIN64A
>> --openssldir=%SystemDrive%\openssl11x64\ && nmake
>> When I add 'no-blake2' OpenSSL compiles fine.
>> Does anyone else have this Issue? Last build I did was at 23rd November
>> and then today. So I am not sure which commit (if at all??) probably
>> introduces this Problem.
> There haven't been any changes to that file since May. AFAICT there also
> haven't been any changes to any of the header files (directly or
> indirectly included) in that time period. Did you update anything on the
> compiler side?
> In any case a compiler hang sounds very much like a compiler bug. No
> matter what we write in our code it should not cause the compiler to
> hang. That said we have made changes in the past to workaround such
> issues if someone is able to figure out exactly what bit of the code it
> is having trouble with.
>> BTW: Unrelated to this: Since 2 or 3 Months, the Fuzz Test always ends
>> with "Error: wstat10 (2560, 0x0)" (Sorry dont remember the full Error -
>> compiling something else right now). All other Tests are ok. Can I keep
>> ignoring this or should I create a BR on Github?
> Please raise this as an issue in github.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 833 bytes
Desc: OpenPGP digital signature
More information about the openssl-users