[Lazarus] Running on mipsel (was: Re: Cross-compiling -> MIPS)

Mark Morgan Lloyd markMLl.lazarus at telemetry.co.uk
Thu Oct 25 16:26:27 CEST 2012


Mark Morgan Lloyd wrote:
> Henry Vermaak wrote:

>> I built lazarus with fpc 2.7.1 for armhf this weekend.  Why do you say
>> 2.6.0 is needed?
> 
> Because of the feeling with which most developers insist that 2.6.0 is 
> the one true way :-)
> 
> OK, I'll try it for MIPS in a couple of days. I presume you were using 
> Lazarus trunk.

Lazarus trunk (r39158) builds OK with 2.7.1 as described in the wiki 
(noting the fairly strict requirement for binutils version etc.), 
although I note warnings like

/usr/bin/ld: /usr/lib/crti.o: warning: linking abicalls files with 
non-abicalls files

on the major programs (lazbuild, lazarus). Lazarus build time (i.e. 
excluding the compiler/RTL/FCL) was about 20 hours, emphasising the 
limitations of Qemu.

IDE startup showed the splash and configuration dialogue, after that it 
appeared to lock up taking minimal CPU, final messages were

SearchDebuggerCandidates Value=gdb File=/usr/bin/gdb
NOTE: help options config file not found - using defaults
TApplication.HandleException Access violation
   Stack trace:
   $01630294  TINDEXCURSOR__VARIANTTOBUFFER,  line 168 of 
./fcl-db/src/dbase/dbf_idxcur.pas
[TMainIDE.Destroy] A
[TMainIDE.Destroy] B  -> inherited Destroy... TMainIDE
[TMainIDE.Destroy] END

until I killed it with a ^C. Restarting resulted in an IDE which 
approximately ran, but was raising that exception every few seconds. It 
wasn't possible to "OK" past this, it would recur and either this or 
something else appeared to be messing up e.g. the IDE's ability to 
store/access ~/.lazarus configuration settings: because of this lazbuild 
wasn't able to build a trivial test project prepared on a different system.

Redoing with a simple  make all  still resulted in approximately similar 
behaviour, but without a reported error line etc.

Unless somebody knows that that revision had issues, I think that some 
more of the compiler's residual code generation issues need to be wrung 
out before trying Lazarus again.

-- 
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]




More information about the Lazarus mailing list