Bug 36495 – Fix compilation on Solaris: mmap and munmap take char*

• Component: JavaScriptCore
• Status: RESOLVED WONTFIX

[Kent Hansen]

"Formal argument 1 of type char* in call to mmap(char*, unsigned, int, int, int, long) is being passed void*"
"Formal argument 1 of type char* in call to munmap(char*, unsigned) is being passed JSC::Register*."
"Invalid static_cast from type char* to type Register*".

[Kent Hansen]

This is a redux of the issues fixed by r34597.

[Kent Hansen]

Created attachment 51431 [details]
Proposed patch

[WebKit Review Bot]

Attachment 51431 [details] did not pass style-queue:

Failed to run "WebKitTools/Scripts/check-webkit-style" exit_code: 1
JavaScriptCore/wtf/TCSystemAlloc.cpp:446:  Weird number of spaces at line-start.  Are you using a 4-space indent?  [whitespace/indent] [3]
JavaScriptCore/interpreter/RegisterFile.h:177:  Missing spaces around |  [whitespace/operators] [3]
Total errors found: 2 in 5 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Kent Hansen]

(In reply to comment #3)
> Attachment 51431 [details] did not pass style-queue:
> 
> Failed to run "WebKitTools/Scripts/check-webkit-style" exit_code: 1
> JavaScriptCore/wtf/TCSystemAlloc.cpp:446:  Weird number of spaces at
> line-start.  Are you using a 4-space indent?  [whitespace/indent] [3]
> JavaScriptCore/interpreter/RegisterFile.h:177:  Missing spaces around | 
> [whitespace/operators] [3]
> Total errors found: 2 in 5 files
> 
> 
> If any of these errors are false positives, please file a bug against
> check-webkit-style.

Just keeping the existing 2-character spacing of that function.

[Darin Adler]

Comment on attachment 51431 [details]
Proposed patch

Can we instead take care of this by setting up a header to overload these functions for void* on Solaris systems? I'd prefer a fix that doesn't require sprinkling the non-standard code around.

Something like this:

    #if OS(SOLARIS)
    static inline void* mmap(void* addr, size_t length, int protections, int flags, int fileDescriptor, off_t offset)
    {
        return mmap(static_cast<char*>(addr), length, protections, flags, fileDescriptor, offset);
    }
    #endif

If we put that in file called <wtf/MManExtras.h> and included that instead of <sys/mman.h> that would fit our normal portability approach.

review- because I'd like you to consider that option

[Kent Hansen]

(In reply to comment #5)
> (From update of attachment 51431 [details])
> Can we instead take care of this by setting up a header to overload these
> functions for void* on Solaris systems? I'd prefer a fix that doesn't require
> sprinkling the non-standard code around.

Hi Darin,
Yep, makes total sense; that should also remove the chance of this issue being brought back to life again by newly introduced calls to mmap/munmap, which is _really_ the problem to solve.

> If we put that in file called <wtf/MManExtras.h> and included that instead of
> <sys/mman.h> that would fit our normal portability approach.

Thanks for the suggestion, I'll try that one.

[Simon Hausmann]

Taking the liberty of closing this as WONTFIX. Neither Kent nor I are working on Solaris support anymore. If anyone would like to tackle this one, feel free to re-open.