View Single Post
Old 01 March 2017, 11:47   #51
Registered User
Join Date: Aug 2007
Location: berlin/germany
Posts: 947
Originally Posted by matthey View Post
The AROS build system is not friendly.
matthey, i beg you not to spread fud. aros build system is _very_ user friendly. even i have no problem to maintain my local builds on different linux contribs. its just sourcing out the necessary packages, configure and make.

it is very easy to work on the different parts of the source to improve it and be able to check the results in an instant. the usage of metamake and the syntax of the makefiles may be a little orthodox, but i bet it was necessary. there always will be some part of an open community driven project one may not particularly like but can live with it.

It uses GCCisms and macros which make it a pain and it requires a fairly new version of GCC which needs special support and has bugs and for the 68k..
may be and may be it needs sophisticated tools, but it is a complex project to maintain as a whole, not a single library replacement.

i compile aros with gcc4.6.4 and 6.3.0 both work well. i have had just one bug fixed lately. the code generation of 6.3.0 has improved in comparison to 3.x.x and 4.x.x. it may yet not be on the level of 2.9.0 and lack amiga extensions, but head over to 6.x.x 68k toolchain thread ( in here to inform yourself more. the patches from that project might even be incorporated into aros backend. one just need to do that.

i really have a feeling that the problems of newer compilers with 68k are partly a myth. few days ago i was presented by phx with a testcase he said it failed on 5.x.x. i compiled it with aros 6.3.0 and the result was correct.

other than that aros is ready for clang/llvm. you may remember that deadwood evaluated vasm as option and phx even did some feature improvements based oh his proposals. as result it occured that there was no practiical avantage to use vasm/vbcc at this point.

The AROS code itself introduced some new concepts, APIs and kludges to the AmigaOS (some of questionable value)
aros wont take over just each and every concept of the genuine os and for a reason (monitor files as example). aros extends os3.x functionality, but isnt it what we want? the extensions may or may not be questionable value. you need to conrtibute something and then you can discuss the options to make it better, simple as it is.

and the code did not need to be as efficient when it was on the x86.
almost no amiga people have given any valid feedback to the developers, and no wonder the 68k developers got disencouraged. most aros developers are x64 oriented, nevertheless as they intoduce the smp and are daily breaking the build they respect 68k as a platform and listen and react to my reports. no wonder though that teir interest isnt optimizing 68k issues at this time.

There is also the problem of some code being unfinished and not ready to be optimized yet. GCC compiler code generation for the 68k does seem to be improving finally after many years of sadness but there is still work to be done to get a cross compiler environment working fully. This alone could make a big difference. I am no expert on AROS.
what? i compile amiga-m68k daily. im not sure of what are you talking about.
wawa is offline  
Page generated in 0.04146 seconds with 10 queries