<html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /></head><body style='font-size: 10pt; font-family: Verdana,Geneva,sans-serif'>
<p>I am seeing something which may be related.</p>
<p>I am seeing failures on a locally built VM in 2 out of 4 machines which fail 2 Cuis FFI unit tests.</p>
<p><span>The FFI tests pass green on aarch64 on Samsung Chromebook Plus/Linux and LePotato/Linux.</span></p>
<p><span>However, in Raspberry Pi 3 [ArchLinuxArm=linux/aarch64] and an old x86_64 Ubuntu Linux machine I get matching failures.</span></p>
<p><span> - FFIPluginTest>>#testGenericDoubleCall2</span><br /><span> - FFIPluginTest>>#testMixedDoublesIntAndStruct</span></p>
<p><br /><span>In each case I am using locally built VMs with the same code from git. My cloned repository (<a href="https://github.com/KenDickey/opensmalltalk-vm">https://github.com/KenDickey/opensmalltalk-vm</a>) was current with OpenSmalltalk's master as of a couple of weeks ago. I have not synched since then as I have a pull request open for my (proposed) changes to support aarch64 and am waiting for integration into the master to test.</span><br /><br /><span></span></p>
<p><span>The other test cases pass.</span><br /><span></span></p>
<p><span>In each case a double result is properly returned (gdb) but I get lost in gcc3x_interp.c floatObjectOf() details.</span></p>
<p><span>I am time stressed at the moment, but gdb is showing that the proper values were calculated by the ffiTest prims.</span></p>
<p><span>Just after the test funs, in the FFIPlugin code (ARM64FFIPlugin.c) there is a code line</span></p>
<p><span> result1 = floatObjectOf(floatRet);<br /></span></p>
<p>Just before this, gdb shows 'floatRet' as correct.</p>
<p><span>Just after, result1 looks good.</span></p>
<p><br /></p>
<p><span>HTH,</span></p>
<p><span>-KenD</span></p>
</body></html>