Re: Release 1.5.9 soon

[Christopher Faylor <cgf at alum dot bu dot edu>]

On Fri, Mar 19, 2004 at 07:41:24PM -0500, Nicholas Wourms wrote:
>vinschen wrote:
>>On Mar 19 15:39, Nicholas Wourms wrote:
>>
>>>...
>>>
>>>>64kb-boundry limitations of MapViewofFile for Win9x/Me/2k/XP:
>>>>http://freespace.virgin.net/james.brown7/tutorials/bigmem01.htm
>>>
>>>Actually, there's a newer version, so this should be:
>>>(Part 1) http://www.catch22.org.uk/tuts/bigmem01.asp
>>>(Part 2) http://www.catch22.org.uk/tuts/bigmem02.asp
>>
>>
>>Hmm, did you read the mmap code in Cygwin already?
>
>No, not in depth...
>*rereads mmap.cc*
>Ah yes, ok, I see the part where you take care of this.  Sorry for the 
>extra noise.
>
>Out of curiosity, how goes the battle with the gcc people?  Are they 
>going to renounce their unportable ways?
>
>Isn't it ironic that they are so overzealous about their precious C++ 
>standards that they are willing break source compatibility with a 1/3 of 
>the code out there, but when it comes to PCH, they are willing to settle 
>for a non-standard, non-portable interface?  Of course, I take it that 
>the mantra around there is that if the OS isn't GPL'ed, who gives a *&%#.

Yes, you're right.  It is disappointing.  IIRC, it isn't just windows
which has problems, either.  And, it's not just gcc that actively
doesn't care.

>And speaking of C++ standards, is it just me or does anyone else find 
>that huge list of features "which used to work, but now cause errors 
>with g++-3.4" totally outrageous?  Do they do it to torture users who 
>have very little interst in C++ semantics and who only want to get a 
>package to compile?  Do they have some keen desire to see developers 
>writhe in agony as they are forced to go through thousands and thousands 
>of lines of code making what are arguably stylistic changes just so g++ 
>will build it?  I don't know, but $20 says g++-3.4 will *refuse* to 
>compile the setup.exe sources.

I was using 3.4 on cygwin sources for my personal builds but I had to
give up due to flakiness.  I'm sure it's better now.  I think I probably
compiled setup.exe at some point, too, but I don't remember exactly.

But, I have to say that, for Red Hat, it isn't a very popular story when
we tell customers "Oh, and in our next release, your previously-built
binaries may not work right and your source code may not compile."  This
is one of the, IMO, big drawbacks of open source.  A company like Red Hat
is basically at the mercy of decisions like this over which they have no
control.  Of course, we could throw people at the problem if we really
wanted to influence decisions, I suppose.

cgf