This is a newly encountered problem when I upgraded to DMD 0.86.  I was driven mad to pinpoint the cause, trimming down one of my programs until there was nothing left of it.  I then finally realised that it wasn't the code it doesn't like, but the filename!

Here's a minimal testcase:

----------
void qwert() {}

int main() { return 0; }
----------

Saved as good_filename.d and bad-filename.d.  Compiler output:

----------
D:\My Documents\Programming\D\Tests\bugs>dmd good_filename.d
Max # of fixups = 4
E:\DMD\BIN\..\..\dm\bin\link.exe good_filename,,,user32+kernel32/noi;

D:\My Documents\Programming\D\Tests\bugs>dmd bad-filename.d
Assertion failure: 'result[i] == '_' || isalnum(result[i])' on line 34 in file 'mangle.c'

abnormal program termination

D:\My Documents\Programming\D\Tests\bugs>
----------

It seems that the compiler is trying to mangle the function name by combining it with the filename.  If I remove the dummy qwert function, the error doesn't show, obviously as main isn't mangled (at least in this way).

Is it legal for D code filenames to have characters other than [_A-Za-z0-9] (besides the . near the end), if they're not used as module names at all?

If so, it should be mangling the name in a way that makes sense.

If not, this should be clearly documented (don't ask me where though), and the compiler should stop with a clean, human-readable error message at the beginning.

Stewart.

-- 
My e-mail is valid but not my primary mailbox, aside from its being the unfortunate victim of intensive mail-bombing at the moment.  Please keep replies on the 'group where everyone may benefit.

I hadn't thought of that, but you're right. The filename characters must only be allowable identifier characters, or the name mangling will fail. This needs to be documented, and a proper error message.

Walter wrote:
> I hadn't thought of that, but you're right. The filename characters must
> only be allowable identifier characters, or the name mangling will fail.
> This needs to be documented, and a proper error message.

Wouldn't it be possible to strip out the bad characters, or replace them with underscores?

Like, if you tried to use "bad-filename.d", it would use "bad_filename.d".  If you tried to use "español.d" it would use "espa_ol.d".  Then again, this is taking a step away from the programmer, and could cause conflicts...

-[Unknown]

"Stewart Gordon" <smjg_1998@yahoo.com> wrote in message news:c786h9$332$1@digitaldaemon.com...
> This is a newly encountered problem when I upgraded to DMD 0.86.  I was driven mad to pinpoint the cause, trimming down one of my programs until there was nothing left of it.  I then finally realised that it wasn't the code it doesn't like, but the filename!
>
> Here's a minimal testcase:
>
> ----------
> void qwert() {}
>
> int main() { return 0; }
> ----------
>
> Saved as good_filename.d and bad-filename.d.  Compiler output:
>
> ----------
> D:\My Documents\Programming\D\Tests\bugs>dmd good_filename.d
> Max # of fixups = 4
> E:\DMD\BIN\..\..\dm\bin\link.exe good_filename,,,user32+kernel32/noi;
>
> D:\My Documents\Programming\D\Tests\bugs>dmd bad-filename.d
> Assertion failure: 'result[i] == '_' || isalnum(result[i])' on line 34
> in file 'mangle.c'

I got the exact same error with 0.86 but then Walter changed something and
i redownloaded it and it now works! My filename also contained '-' but
i didn't figure out this was the problem.
I was having this problem with the version where dmd.exe
is around 2MB, but in the new one it is back to its usual size.

> abnormal program termination
>
> D:\My Documents\Programming\D\Tests\bugs>
> ----------
>
> It seems that the compiler is trying to mangle the function name by combining it with the filename.  If I remove the dummy qwert function, the error doesn't show, obviously as main isn't mangled (at least in this way).
>
> Is it legal for D code filenames to have characters other than [_A-Za-z0-9] (besides the . near the end), if they're not used as module names at all?
>
> If so, it should be mangling the name in a way that makes sense.
>
> If not, this should be clearly documented (don't ask me where though), and the compiler should stop with a clean, human-readable error message at the beginning.
>
> Stewart.
>
> --
> My e-mail is valid but not my primary mailbox, aside from its being the unfortunate victim of intensive mail-bombing at the moment.  Please keep replies on the 'group where everyone may benefit.

"Unknown W. Brackets" <unknown@at.simplemachines.dot.org> wrote in message news:c78tos$18hg$1@digitaldaemon.com...
> Walter wrote:
> > I hadn't thought of that, but you're right. The filename characters must only be allowable identifier characters, or the name mangling will fail. This needs to be documented, and a proper error message.
>
> Wouldn't it be possible to strip out the bad characters, or replace them with underscores?
>
> Like, if you tried to use "bad-filename.d", it would use "bad_filename.d".  If you tried to use "español.d" it would use "espa_ol.d".  Then again, this is taking a step away from the programmer, and could cause conflicts...

I think that such would cause more grief, like how Microsoft mangles long filenames into 8.3 ones.

Stewart Gordon wrote:

> This is a newly encountered problem when I upgraded to DMD 0.86.  I was driven mad to pinpoint the cause, trimming down one of my programs until there was nothing left of it.  I then finally realised that it wasn't the code it doesn't like, but the filename!
<snip>

I upgraded to 0.88 and it just stopped complaining!  But still....

Stewart.

-- 
My e-mail is valid but not my primary mailbox, aside from its being the unfortunate victim of intensive mail-bombing at the moment.  Please keep replies on the 'group where everyone may benefit.