[STDCXX-1045] [OpenSolaris] compilation errors on madvise

• Type: Bug
• Status: In Progress
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 4.2.1
• Fix Version/s: 4.2.2
• Component/s: Build and Installation
• Severity: Compiler Error
• Assignee: Martin Sebor
• Reporter: Martin Sebor
• Created: 18/Jan/10 01:00
• Updated: 18/Jan/10 01:03

Description

From Ben's email below it sounds as though we need a test for posix_madvise and use it in favor of alternatives such as madvise when it's available:

-------- Original Message --------
Subject: madvise on Solaris
Date: Fri, 18 Dec 2009 18:00:30 +0000
From: John Taylor <[e-mail]>
To: msebor AT gmail.com

[...]
Anyway, I build stdcxx-4.2.1 on Solaris 10 and SXCE.

I notice that madvise doesn't get recognized, due to how the
function is defined. Not a problem. Opensolaris, and late
versions of SXCE, have introduced a function wrapper called
posix_madvise, which basically wraps the madvise function
so that it's more palletable to the C++ compiler.

In line 134 of src/memattr.cpp, there's a comment that I don't really
understand,
nor can I find any reference to the "unreliable" comment via google or stdcxx
lists.

#  ifdef _RWSTD_OS_SUNOS

        char dummy = '\0';

        // on Solaris use mincore() instead of madvise() since
        // the latter is unreliable

I admit to not being a very experience c/c++ programmer, but from my
read of the solaris man pages for mincore and madvise, I'm not
entirely convinced that this is a better work around. Does anyone really
know what this "unreliable" means?

Would it be appropriate to "duplicate" Sun's wrapper for S10 and
SXCE/OSOL which don't have posix_madvise (given it's CDDL license
status) and use posix_madvise where approrpriate?