[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] compile error



Oh you make it sound like it's easy but it's not or else it'd be done.  I see people package one or two things, but nobody pushing 100 updated packages except Alan and Vincent, two generous nuts with lots of time ;)  Some of the patches for our system are NOT simple and all of those patches have to be ported to the new software.  Back in the day Keith and I were fighting with and wasting a lot of time on gcc and binutils.  Thanks to vincent we don't have to do that anymore.

Much of the patches Alan has done on gentoo may directly port over.  I hope the ebuilds are commented.

Most new software you can pull down the spec files from their packages but they won't build without a newer version of rpm which won't build for various reasons (not counting mmap problem).  If you don't use the newer spec, you have to tailor the spec of each old rpm to the new build output and while not incredibly complicated is incredibly time consuming.  Since the old software won't build with the current gcc you must be perform this spec update for basically every package.  One error and the build fails and you have to start over.  One instance of malloc bug and you have to start over.  This was an infuriating process designed by satan himself ;)  There are mintlib deficiencies to work around and now every use of -lsocket has to be reverted.  Lack of dynamic libraries create various issues that you have to learn about in each package and work around.

Plus I was trying to establish build vs runtime dependencies so that an autobuilder could be made.  This isn't necessary to just "rebuild sparemint" but it's necessary to do it right and have the process become relatively automatic.

Yes we maybe will be able to get by without mmap or other things but it's a huge amount of work that isn't worthwhile.  It's better to do it right and do what needs to be done to modernize the OS.

I go to work every day and the mentality is "get it done, I don't care if it's half assed and broken".  When I work on my hobby I will choose to do it right, and if it takes 5 years because I  have real life problems and a 1 year old kid, then someone else can feel free to step in.  I don't mind someone making real progress in a half-assed way but if I do something right and someone ruins it, that will upset me and I'll quit. 

Luckily I've gotten almost nothing done so nothing to worry about :-P

At this point you should simply use what is actively supported - gentoo.  If I manage to get more time (I might be able to soon) and I can properly support sparemint things could get better.  But I am understand packaging and I don't understand OS internals and assembler so on a hobbyist level that is what I want to work on with my limited time.

Thanks,
Mark




On 2/8/2013 1:55 PM, Miro Kropáček wrote:

On Fri, Feb 8, 2013 at 6:55 PM, Mark Duckworth <mduckworth@atari-source.org> wrote:
Attempts to update result in various difficulties like lack of mmap().  Not sure how Alan is working around that, just renaming calls to malloc (which would work) or what.
Whoa, wait. Lack of mmap() is a problem maybe in 1% of all software which is available (I can think of git and... yes, the rpm). All the normal stuff, starting from 'bash' and ending with 'python' can be build using normal ./configure --host=m68k-atari-mint && make && make install. You can see that in Alan's releases and also in Vincent's unofficial builds. It's really mainly matter of packaging.


--
MiKRO / Mystic Bytes
http://mikro.atari.org