Well now I can't even compile the example which used to crash at runtime: http://codepad.org/Dh9fjQKG

This is while trying to compile the project in the attachment in this ticket: http://www.dsource.org/projects/qtd/ticket/85

Btw that's using 2.054 on XP32.

Also I don't understand this, see changeset 411:

> hg log -r :

changeset:   411:6fe37cbd9510
user:        David Nadlinger <code@klickverbot.at>
date:        Mon Apr 18 00:16:33 2011 +0200
summary:     Octal literals have been deprecated, fix build.

changeset:   412:4189f1d51a7d
user:        Max Samukha <maxsamukha@gmail.com>
date:        Thu May 05 20:11:26 2011 +0300
summary:     support for new dmd naming

changeset:   413:bdc08c8391ad
user:        Max Samukha <maxsamukha@gmail.com>
date:        Thu May 05 22:32:50 2011 +0300
summary:     removed unnecessary regex grouping

changeset:   414:b2a803c73b89
tag:         tip
user:        David Nadlinger <code@klickverbot.at>
date:        Fri May 06 13:39:49 2011 +0200
summary:     Declare tabArray const.

So now I'm at 414, and when I try to build: D:\dev\projects\qtd\examples\desktop\systray>build

D:\dev\projects\qtd\examples\desktop\systray>..\..\..\tools\drcc\drcc systray.qrc  1>qrc_systray.d

D:\dev\projects\qtd\examples\desktop\systray>dmd -unittest
-version=QtdUnittest -ID:\dev\lib\D\qtd\build_dir\build
-ID:\dev\lib\D\qtd\d2 ..\..\..\build_dir\build\lib\qtdcore.lib
..\..\..\build_dir\build\lib\qtdgui.lib qrc_systray.d
window.d main.d -ofsystray
D:\dev\lib\D\qtd\d2\qtd\ctfe\Integer.d(157): octal literals 0123 are
deprecated, use std.conv.octal!123 instead
D:\dev\lib\D\qtd\d2\qtd\ctfe\Integer.d(164): octal literals 0123 are
deprecated, use std.conv.octal!123 instead

Are you only testing *D1* versions of QtD? Because that could explain why you're apps are running while mine are not. :)

s/you're/your

Wait wait wait. Sorry this is completely my fault the build script copied old directories. Sorry for the noise!

Ok, but the other CTFE issues are still there:

D:\dev\projects\qtd\examples\desktop\systray>dmd -unittest
-version=QtdUnittest -I..\..\..\build_dir\build -I..\..\..\d2
..\..\..\build_dir\build\lib\qtdcore.lib
..\..\..\build_dir\build\lib\qtdgui.lib qrc_systray.d window.d main.d
-ofsystray
..\..\..\d2\qtd\MOC.d(271): Error: gen.output ~= format_ctfe("\x0a //
content:\x0a") cannot be evaluated
at compile time
..\..\..\d2\qtd\MOC.d(272): Error: cannot evaluate format_ctfe("
${},       // revision\x0a",2) at compile time
..\..\..\d2\qtd\MOC.d(484): Error: cannot evaluate generateCode(gen)
at compile time
..\..\..\d2\qtd\MOC.d(513): Error: cannot evaluate generateMetaInfo()
at compile time
..\..\..\d2\qtd\MOC.d(513): Error: argument to mixin must be a string,
not (generateMetaInfo())

On 7/21/11 2:14 AM, Andrej Mitrovic wrote:
> Ok, but the other CTFE issues are still there:
>
> D:\dev\projects\qtd\examples\desktop\systray>dmd -unittest
> -version=QtdUnittest -I..\..\..\build_dir\build -I..\..\..\d2
> ..\..\..\build_dir\build\lib\qtdcore.lib
> ..\..\..\build_dir\build\lib\qtdgui.lib qrc_systray.d window.d main.d
> -ofsystray
> ..\..\..\d2\qtd\MOC.d(271): Error: gen.output ~= format_ctfe("\x0a //
> content:\x0a") cannot be evaluated
> at compile time
> ..\..\..\d2\qtd\MOC.d(272): Error: cannot evaluate format_ctfe("
> ${},       // revision\x0a",2) at compile time
> ..\..\..\d2\qtd\MOC.d(484): Error: cannot evaluate generateCode(gen)
> at compile time
> ..\..\..\d2\qtd\MOC.d(513): Error: cannot evaluate generateMetaInfo()
> at compile time
> ..\..\..\d2\qtd\MOC.d(513): Error: argument to mixin must be a string,
> not (generateMetaInfo())

Well, seems like I never built one of my applications since the CTFE overhaul was merged into master… ;) It would be great if you could have a short look at what actually got broken. qtd.MOC is hand-written, so tracking that CTFE regression [1] down without going insane should be possible. By the way, pretty much all of my QtD code is D2.

If you don't have time for it, I will definitely have a look at it at some point in the future, but after my Thrift project, updating LDC 2 is currently on top of my todo list.

Thanks,
David


[1] From a DMD development perspective, it might be that this isn't a regressions because the feature might not have worked correctly before, but it certainly did the job. If I were to make a wild guess, I'd try manually setting gen.output to [] if it should be null at that point, because I had a similar problem after the overhaul with strings, but as this is supposed to be fixed, it's probably something else.