Well, to put my two cents in on this.... one of the reasons my
organization didn't move forward from 3.x to 4.x in the first place was
the short period of time when it was easy to perceive that the AOLServer
community wasn't interested in Windows builds at all.  That was a
misperception, but it was a critically timed one (we'd just replaced the
entire development team).

We do our 3.4.x builds using Visual Studio 2003 and have long hacked on
the .vcproj files to get them to do what we need. It's often painful to
remember that adding a new file to the build requires a fix to those
files (or to the Unix side). And we can never keep the "debug" vs.
"release" stuff straight; those files really are painful.

We like to use the Microsoft compilers (and the VS suite) for a few
reasons:
* Our customers trust them.
* Compiler diversity. Our modules are written in C++. We find it very
useful to compile on at least three different vendors just to get their
warning messages.  To be honest, my organization could have given up on
Solaris builds on the commercial side (nobody's bought it in years), but
I keep those builds going (using the Sunsoft compilers) just for
engineering quality checks.  Getting stuck with just GCC would be a real
problem on this front.
* We like to use Microsoft DLLs when they're available for the job we're
solving.  For example, our crash dump libraries are written specifically
to their debugger support library.
* Debugging

This goes to a part of the M$ development environment that's also
useful: their debugging environment (even the free-as-in-beer windbg)
sometimes proves better at finding the kinds of race conditions &
locking problems we seem to write once a year and chase for months. It's
certainly better at working with customer sites: we can usually verify
whether a customer crash is a duplicate of something in our lab using
their "minidumps", which will come in much smaller, since they only
store stacks.  I don't know if this sort of scheme is available in the
MSYS world. (Of course, we do deal with the GB or bigger crashes in
Linux -- and if the crash *isn't* a duplicate, we have to generate a
full dump somewhere in Windows. The win is in being able to say "we know
this one and we're on it" a lot faster).

Just trying to provide insight; we haven't even downloaded an AOLServer
4 build in a year, and I don't know when we'll be able to try it.

-- ReC

-----Original Message-----
From: AOLserver Discussion [mailto:[EMAIL PROTECTED] On Behalf
Of Andrew Piskorski
Sent: Sunday, July 02, 2006 1:48 PM
To: [email protected]
Subject: Re: [AOLSERVER] Unix vs. Windows build environments

On Fri, Jun 30, 2006 at 09:52:16AM -0400, Rick Gutleber wrote:

> I do know that VC++ 2003 was, at least at the time, one of the best 
> compilers in the Windows platform (except for maybe Intel's), and 
> definitely the most standards-compliant.  Yes, from Microsoft.

Then that's yet another good reason to try to retain VC++ support for
AOLserver, if feasible.

> Of course, there's always gcc, right?  I think providing a MS project 
> file is highly desirable (VS6 is still widely used despite being as
old 
> Windows98 and newer versions of VS can handle .dsw/.dsp files), but 

On that bit I disagree.  I've used those VC++ 6.0 project files, and
they really suck.  They are better than nothing at all, but that's
about it.  I don't think VC++ 6.0 project files do a good job serving
the needs of ANY AOLserver user.  No matter who you are, one or more
other solutions would be a lot better - if they're supported by the
AOLserver community.

I assert that all people who have any desire to compile AOLserver on
MS Windows fall into at least one of these three categories:

1. I don't want to compile anything, I just want to use stock
AOLserver and all its modules on MS Windows, but compiling them myself
seems to be the only way I can do that!  Ugh, I hate this, but I am
persistent, maybe I can figure it out...

2. I'm building stock AOLserver on MS Windows so Guy #1 above can use
it.  Hm, these build tools look pretty ugly.  Maybe I'll hack
something else together for my own use, but it'll probably never make
it back into the official AOLserver build process.

3. I am doing custom development using AOLserver, either just on
Windows or cross-platform, and both Unix and Windows.  Most of my C
code is in my own custom AOLserver modules, but I need to be able to
patch the core code if necessary too.  I MUST be able to build all of
AOLserver from scratch myself, using tools I am happy with.

Guy #1 above just wants working AOLserver binaries, all modules, all
versions, automatically built, easily installed, ready to go.  AKA, at
least as far as the C code goes, he is a customer, not a developer.
He doesn't much care what the build tools are, as long as they work
out of the box, for everything.

Guy #2 would mostly be happy with anything Guy #3 below is happy with,
perhaps biased more towards simplicity rather than flexibility and
power.  However, he does not want to buy any Microsoft compilers or
GUIs that he doesn't already own, not just for this.

Guy #3 does NOT want ancient kludgy VC++ 6.0 project files.  He might
be more or less willing to deal with them if they're the only option,
but in no case does he really want them.  I think Guy #3 comes in two
major flavors, let's call these Guy #3u and #3w:

Guy #3u is a Unix geek crossing over to Windows.  He wants some GOOD
programmtic build evironment.  Although far from ideal, configure and
make would be ok, *if* they actually work on Windows.  Using a zero
cost compiler is often pretty important, and Microsoft's compiler is
preferred because that's the compiler everyone else on Windows seems
to use.  He MIGHT be willing to buy and use Microsoft's GUI build
tools if that route is already supported, and if using them isn't
awful; then again, he might not.

Guy #3w is a Windows geek (but one who is interested in using
AOLserver, remember).  He'd probably prefer whatever Microsoft's
default GUI-ish build environment is in their latest VC++ 2003 or 2005
compiler toolkits.  However, he'd also be happy with whatever Guy #3u
above is happy with, AS LONG as it isn't a totally baroque Cygwin-like
toolkit.  Microsoft compiler is VERY highly desired.  He MIGHT be
willing to use gcc, Mysys, configure, and make on Windows, just
barely; then again, he might not.  (Configure and make are fairly ugly
even on Unix, using using them on Windows can only make them more so.)

I've pulled the above scenarios out of the air, and have absolutely
zero scientific evidence that they're accurate.  I suspect though that
they do have some truth to them.

I myself fit the Guy #3u profile above.  I am no Windows geek, so the
#3w profile is particularly vague and built from guesswork.  Perhaps
Jamie R. or others could correct or clarify it.  :)

-- 
Andrew Piskorski <[EMAIL PROTECTED]>
http://www.piskorski.com/


--
AOLserver - http://www.aolserver.com/

To Remove yourself from this list, simply send an email to
<[EMAIL PROTECTED]> with the
body of "SIGNOFF AOLSERVER" in the email message. You can leave the
Subject: field of your email blank.


--
AOLserver - http://www.aolserver.com/

To Remove yourself from this list, simply send an email to <[EMAIL PROTECTED]> 
with the
body of "SIGNOFF AOLSERVER" in the email message. You can leave the Subject: 
field of your email blank.

Reply via email to