disabled c99 in Solaris builds

[Magnus Ihse Bursie]

On 2018-12-12 09:24, Baesken, Matthias wrote:
>> FWIW, in my in-development patch set for JEP 347
>> (https://bugs.openjdk.java.net/browse/JDK-8208089)
>>
>> the -xc99=%none% option has been removed and -std=c99 added (for C
>> code). (C++14 includes C99 by reference, so I made that change for
>> consistency in case there were any ABI differences.)
>>
> Hello Kim , do you have an estimation when your patch will be in  ?
>
> Or would you suggest to do a separate change and bring in into  jdk/jdk (and  jdk12) already?
I'd recommend a separate change to remove the -xc99=%none%. That's 
trivial and can be done immediately, to the benefit of everyone. Getting 
C++14 working is not happening quickly.

/Magnus
>
> Best regards, Matthias
>
>
>> -----Original Message-----
>> From: David Holmes <david.holmes at oracle.com>
>> Sent: Dienstag, 11. Dezember 2018 23:16
>> To: Erik Joelsson <erik.joelsson at oracle.com>; Baesken, Matthias
>> <matthias.baesken at sap.com>; 'build-dev at openjdk.java.net' <build-
>> dev at openjdk.java.net>
>> Subject: Re: disabled c99 in Solaris builds
>>
>> On 12/12/2018 3:27 am, Erik Joelsson wrote:
>>> Hello,
>>>
>>> I do not know why this flag was introduced, but it has been there for a
>>> long time. In JDK7 it's listed in jdk/make/common/Defs-solaris.gmk:
>>>
>>> #       -xc99=%none     Do NOT allow for c99 extensions to be used.
>>> #                       e.g. declarations must precede statements
>>>
>>> and was there since the first mercurial change.
>> I can reasonably imagine that this was added to prevent introducing
>> shared code, developed on Solaris, that would not compile on Windows.
>> But those days are long gone.
>>
>> I was bitten by this just this week when an enum declaration compiled
>> fine everywhere but Solaris!
>>
>>> I personally wouldn't mind ditching it.
>> +1
>>
>> David
>>
>>> /Erik
>>>
>>> On 2018-12-11 08:17, Baesken, Matthias wrote:
>>>> Hello , it  seems that currently the Solaris  Oracle Studio Build
>>>> environment is the only one  that explicitly
>>>> forbids C99 C code  by setting  -xc99=%none .
>>>> The current Linux/Mac/AIX/Windows build envs had no issues with the
>>>> coding.
>>>>
>>>> For example I was running into an error with the C variable
>>>> declaration order issue (small example below) today in my coding.
>>>> Is this still a wanted behavior ? What was the reason behind setting
>>>> -xc99=%none , and is  the reason  still valid ?
>>>> I remember we had issues with C99 compatibility back then when VS2010
>>>> was used on Windows, but I think these  days we use VS2013+, is this
>>>> correct ?
>>>>
>>>> The example program  mixes declarations and "other statements" ,
>>>> which needs C99,  I compile with Oracle Studio 12u4 .
>>>>
>>>> /compiler/SS12u4-Oct2017/SUNWspro/bin/cc  vardecl.c -o vardecl
>>>>
>>>> No settings -> works nicely
>>>>
>>>>
>>>> - with C99 disabled as OpenJDK does :
>>>> ----------------------------------------------------------
>>>>
>>>> /compiler/SS12u4-Oct2017/SUNWspro/bin/cc -xc99=%none   vardecl.c -o
>>>> vardecl
>>>> "vardecl.c", line 8: warning: declaration can not follow a statement
>>>>
>>>>
>>>> - with C99 disabled + errwarn as OpenJDK does :
>>>> ------------------------------------------------------------------------
>>>>
>>>> /compiler/SS12u4-Oct2017/SUNWspro/bin/cc -xc99=%none  -
>> errwarn=%all
>>>> vardecl.c -o vardecl
>>>> "vardecl.c", line 8: declaration can not follow a statement
>>>> cc: acomp failed for vardecl.c
>>>>
>>>> example program :
>>>> ---------------------------------------
>>>>
>>>> bash-3.2$ more vardecl.c
>>>> #include <stdio.h>
>>>>
>>>> int main(void) {
>>>>     int a = 0;
>>>>     printf("a: %d \n", a);
>>>>
>>>>     int b = 1;
>>>>     printf("b: %d \n", b);
>>>>     return 0;
>>>> }
>>>>
>>>>
>>>> Best regards, Matthias