[fpc-devel] Error building branch targetandroid
Thomas Schatzl
tom_at_work at gmx.at
Sun Jul 8 14:42:44 CEST 2012
Hi,
Am 08.07.2012 14:13, schrieb Felipe Monteiro de Carvalho:
> On Sun, Jul 8, 2012 at 1:19 AM, Thomas Schatzl <tom_at_work at gmx.at>
> wrote:
>> Updated/rebased the branch with 21814. The sources now compile with
>> -Cparmv6, and I also got cross- and native arm/android compilers.
>
> Great! I will test now.
>
> Just to know: Does this compiler contain all the fixes from multiple
> bug reports including fixes for threading, dynamic library loading,
> etc?
Not sure what exactly you're referring too, but it is on the level of
yesterday evenings' trunk, and does include some of the android specific
changes from the bug tracker, at least those that I deemed correct.
- Libraries are working: note that envp/argc/argv contain nil values.
(first import)
- there were some fixes to threading (r21066)
- some other, minor fixes (r21065)
As far as testing goes, I could package a crosscompiled shared library
into an Android apk and run it (some sample), and successfully started
the compiler natively (displaying the help page); not with the current
revision yet though..
>> Note that you may need to mark your android app as conforming to
>> armeabi-v7
>> now because the compiler may emit or at least allows armv6
>> instructions
>> (whatever these are) now.
>
> That would be counter-productive because my HTC Wildfire runs ARMv6
> so
> my application would stop working there.
>
> ARMv5 Android devices are virtually inexistent. In my office we have
> 100+ devices and I am not aware of a single one which uses ARMv5. All
> low-end devices are ARMv6
It's Google that defined those ABI levels, not me.
>> Actually, I saw that armv5 already supports blx, do not know right
>> now why
>> it isn't generated in the first place. Did you ever try compiling
>> the
>> compiler or your program with -Cparmv5? Maybe it's just that the
>> compiler
>> default is < armv5?
>
> I never really tested -Cparmv5, only -Cparmv6 and with no options (so
> default value), I just conected the points based on your assertion
> that default=v5, but indeed maybe it is v4.
... or even to some lower common denominator.
Thomas
More information about the fpc-devel
mailing list