Failed to build with dmd2.056

[SamHu]

C:\person\dlang\he-the-great-Juno-Windows-Class-Library-d003eda>set PATH=c:\pers
on\dmd2\windows\bin;c:\person\dmd2\windows\lib;c:\person\dmd2\import;C:\OraHome1
\bin;C:\Program Files\Oracle\jre\1.1.7\bin;C:\Windows\system32;C:\Windows;C:\Win
dows\System32\Wbem;C:\Windows\System32\WindowsPowerShell\v1.0;C:\Program Files
Microsoft Application Virtualization Client;C:\Program Files\NetManage\RUMBA\Sys
tem;C:\Program Files\NetManage\RUMBA;C:\Program Files\QuickTime\QTSystem;C:\Pr
ogram Files\Intel\DMIX;F:\Apps\INFORMX\Bin;

C:\person\dlang\he-the-great-Juno-Windows-Class-Library-d003eda>dmd @files.rsp -
O -release -inline -lib -ofjuno.lib
Assertion failure: '!newval || isStackValueValid(newval) || isRefValueValid(newv
al)' on line 5837 in file 'interpret.c'

abnormal program termination

[JesseKPhillips]

Check the makefile. It uses juno.args (though I probably should just update files.rsp), but it also compiles two files separately before building juno.lib. There seems to be a compiler/linker bug when building them all together.

[Numpsy]

fwiw i just built DMD2 and Phobos from the current git code, and this issue doesn't seem to happen with those.

[Numpsy]

actually, it seems that the above command line with DMD 2.057 and the latest Juno fails with

object.di(440): Error: variable __aatmp1604 used before set

works ok without -inline.
Nice :-(

[Numpsy]

Looks like the error comes from xml/xsl.d and is related to iterating over the VARIANT[XmlQualifiedName] in XsltArgumentList.

Seems to be related to VARIANT.opAssign, and adding braces around the call to clear() at the top of it makes the error go away (i'm not seeing why it should though).

[JesseKPhillips]

I've updated the make file I'm using so that these options are passed to the compiler and separated xsl.d from the inclusive build so that it will compile with these options.

I'll leave this open as it would be good to file a bug report if it can be reduced.

[Numpsy]

I've just tried to build with the latest DMD/Phobos Git, and i now get what appears to be Phobos bug 7199/4841.
Still, at least something is changing :-(