• BCS (25/26) Mar 13 2008 I was compiling the D grammar using my parser generator and (after getti...
• Russell Lewis (9/46) Mar 17 2008 I have been seeing some SEGVs occasionally in my template-rich program,
• BCS (2/12) Mar 17 2008 BTW I don't what I found is a bug, I think it counts as abuse of DMD!

BCS <ao pathlink.com> writes:

I was compiling the D grammar using my parser generator and (after getting 
by the 3 page long error messages) I was hopeful that it would build. However 
DMD then seg-v'ed.

transcript:

[bshrop localhost trunk]$ time dmd syntax/dparse.d -debug=dparse_unittest -J.
     glue/templates.d



name: Foo
name: Foo
name: StaticConstructor
name: StaticDestructor

... more of the same ...

Directly recursive rule: OrExpression
Directly recursive rule: AndAndExpression
Directly recursive rule: OrOrExpression
Directly recursive rule: AsmBrExp
Segmentation fault

real    6m36.121s
user    4m47.630s
sys     0m21.190s
[bshrop localhost trunk]$ ls


here is a screen shot of what I think is the issue (sorry I couldn't get 
that as text as well)

http://www.uidaho.edu/~shro8822/WOW.JPG

I think DMD has a bug in it when it tries to dump an object file after using 
up 2.25GB of ram

(I don't think this one needs to go into the bugzilla <G>)

Russell Lewis <webmaster villagersonline.com> writes:

I have been seeing some SEGVs occasionally in my template-rich program, 
too.  I haven't tracked 'em down enough for a bug report, yet, but at 
least one (where DMD took up GBs, like your example) turned out to be a 
recursive template in my code.  I found it and fixed it without doing 
the due dilligence of snipping the code enough to post a DMD bug report. :(

However, I have other SEGVs (one is hitting me right now) where DMD 
doesn't consume much memory.  Gonna have to actually file this one, once 
I get it isolated.

BCS wrote:
 I was compiling the D grammar using my parser generator and (after 
 getting by the 3 page long error messages) I was hopeful that it would 
 build. However DMD then seg-v'ed.
 
 transcript:
 
 [bshrop localhost trunk]$ time dmd syntax/dparse.d 
 -debug=dparse_unittest -J.
     glue/templates.d



 name: Foo
 name: Foo
 name: StaticConstructor
 name: StaticDestructor
 
 ... more of the same ...
 
 Directly recursive rule: OrExpression
 Directly recursive rule: AndAndExpression
 Directly recursive rule: OrOrExpression
 Directly recursive rule: AsmBrExp
 Segmentation fault
 
 real    6m36.121s
 user    4m47.630s
 sys     0m21.190s
 [bshrop localhost trunk]$ ls
 
 
 here is a screen shot of what I think is the issue (sorry I couldn't get 
 that as text as well)
 
 http://www.uidaho.edu/~shro8822/WOW.JPG
 
 I think DMD has a bug in it when it tries to dump an object file after 
 using up 2.25GB of ram
 
 (I don't think this one needs to go into the bugzilla <G>)

BCS <BCS pathlink.com> writes:

Russell Lewis wrote:
 I have been seeing some SEGVs occasionally in my template-rich program, 
 too.  I haven't tracked 'em down enough for a bug report, yet, but at 
 least one (where DMD took up GBs, like your example) turned out to be a 
 recursive template in my code.  I found it and fixed it without doing 
 the due dilligence of snipping the code enough to post a DMD bug report. :(
 
 However, I have other SEGVs (one is hitting me right now) where DMD 
 doesn't consume much memory.  Gonna have to actually file this one, once 
 I get it isolated.
 


BTW I don't what I found is a bug, I think it counts as abuse of DMD!