[Ecls-list] OS X Build problem

[Juan Jose Garcia-Ripoll]

2006/12/10, Karsten Poeck <Karsten.Poeck at gmail.com>:
> I remember errors compiling cl-http with msvc because of the size of the
> generated c-code. With mingw it worked just fine.

I am not sure about the size of code. That is something I should
check. It probably depends very much on the optimization options. I
know GCC can be rather aggresive, inlining functions in the same file.

> I also suspect that mingw is much faster than the msvc-version but still
> have to verify this.

No, this is not true. ECL compiles with MSVC in five minutes. I can
barely distinguish the output on the monitor, the compilation is so
fast. Building with mingw takes, from my old memories, four times or
more -- I mean _discounting_ the configuration process which takes
ages.

> Don't know if the cygwin port is too usefull, since the other two exist, but
> I would love to be able to use the mingw port.
> Is there somethinh we could do to help with the mingw problem (e.g. by
> tracking down a stable version?)

Really, the only solutions I see are getting a vmware image with
everything preinstalled or trying with other versions of Windows
(which I can not for the moment). It seems that mingw and the newest
XP Home SP2 do not get well together. As I said in the previous email,
I have tried all combinations of mingw, msys and msys-dtk

My guess is that something is deeply broken with Mingw: for instance,
msys.dll as shipped with all downloadable packages just crashes. The
problem, according to the developers, has to be tracked to cygwin
(they seem to share some code) and their solution has been to release
a DLL which, when that bug happens, does not produce an error message.

I just wonder whether both mingw and cygwin try to implement their own
DLL loading mechanism and that simply crashes due to the most recent
security mechanisms in Windows (nonexecutable memory, SP2, antivirus,
etc). Note that my hardware is Core 2 Duo, which is a different beast.
I did not have these problems with older machines.

Help is of course welcome, but this seems to be one of those "works
for everybody else" problems :-/

Juanjo

-- 
Dpto. de Fisica Teorica I, Fac. de CC Fisicas, Universidad Complutense,
Ciudad Universitaria s/n Madrid 2804 (Spain)
http://teorica.fis.ucm.es/~jjgarcia/