> Do you really get a gcc error about -WBsymbolic?
> BTW: According to Google -WBsymbolic is apparently exclusively used by
> Neko. A bit strange. Maybe it's a typo? I see there is -Bsymbolic. But
> that should work on 64bit...
Yes, that's how I've discovered that options is not supported. I would
have no idea otherwise. :D -symbolic in any incarnation is not
recognized by GCC 4.7. I found some mention of it at a mailing list
from few years ago, where the person describing this option said that
it is only supported on certain platforms, and amd64 was not listed
there. But I don't even know what the option means. man gcc says this
about it:
-symbolic
Bind references to global symbols when building a shared
object. Warn about any unresolved references
(unless overridden by the link editor option -Xlinker -z
-Xlinker defs). Only a few systems support
this option.
Which sounds like a friendly AI exercising its skills in completing
ambiguous sentences in natural language.
>
> What error did you get there?
At the very start it couldn't find any include, so I started looking
for what may possibly be missing. Eventually, first missing include
was the one from Apache, and so I did yum install apache2-devel or
whichever was required - and woo-hoo, I've got a version that cannot
compile with HaXe. The difference, so far I could understand is in
some Apache-runtime thing... please don't ask me what it is, all I
know it is a header with some cryptic three-letter name, which I
immediately forgot. The thing about it, Apache decided to dedicate a
special project to that runtime. It used to be included with httpd
project, but it is no more. So, the includes in HaXe directories
"think" that Apache runtime is in the same directory as httpd sources,
but in actuality it's somewhere else.
>
> Are Fedora include-directories that different from other Linuxes?
I think they are typical of RHEL, must be very similar to RHEL as this
is a community edition of it. But the problem is that Debian is very
far behind RHEL in a sense of versions of the packages it provides. So
without ever really wanting to, I'm way ahead of the Debian users :)
using all new and shiny kernel 3.5 and other very advanced
technologies... with the downside being, well, complete disarray :)
>
> What tells you that you have newer headers? Again: What error did you
> get there?
See above, Apache runtime vs httpd.
>
> Fedora 17 you have? I might install it on a virtual machine for a try...
>
Yup, would be cool if you do. Also, take a look here, if you will get
to it:
http://pastebin.com/hx1sTZ6M this was my attempt at patching
NME installer script so that it would fit both RHEL and Debain. I
think it's almost there already.
Most interestingly is that there exists a version of nekovm
installable and functioning on Fedora - so it must be possible to
compile it (I can get it with yum). But it will not work with NME -
will try to load 32-bit binaries for some reason. Also, if you are
unfamiliar with typical building / installation process on RHEL, it
would be done with yum-builddep tool. It is supposed to be "smart"
enough to understand the makefile and to download the required
dependencies etc. So "ideally" if anyone ever will design a workflow
for building nekovm on that platform, then, it is good to use that
tool; or at least that would be a good indicator of the typical
problems a user like myself will encounter :)
Best.
Oleg