[Issue 3719] New: forward references can cause out-of-memory error

d-bugmail at puremagic.com d-bugmail at puremagic.com
Mon Jan 18 11:29:01 PST 2010


http://d.puremagic.com/issues/show_bug.cgi?id=3719

           Summary: forward references can cause out-of-memory error
           Product: D
           Version: 2.039
          Platform: Other
        OS/Version: Windows
            Status: NEW
          Keywords: patch
          Severity: normal
          Priority: P2
         Component: DMD
        AssignedTo: nobody at puremagic.com
        ReportedBy: r.sagitario at gmx.de


--- Comment #0 from Rainer Schuetze <r.sagitario at gmx.de> 2010-01-18 11:29:00 PST ---
If there are forward references that are not resolved for a long time, dmd can
waste a lot of memory and cpu-cycles. 

It's not easy to create a small test case, but qtd fails to compile with "out
of memory" without the following patch:

Module::runDeferredSemantic() should not retry between every new symbol in a
class or struct, but only when some sensible progress is made (e.g. if
Module::dprogress > 0). runDeferredSemantic() must then be forced to run once
before running semantic2() on the modules.


Index: mars.c
===================================================================
--- mars.c    (revision 336)
+++ mars.c    (working copy)
@@ -1109,6 +1109,12 @@
     if (global.errors)
     fatal();

+    if(modules.dim)
+    {   // last chance to resolve deferred symbols
+    Module::dprogress = 1;
+    ((Module *)modules.data[0])->runDeferredSemantic(); // should be static
instead
+    }
+
     // Do pass 2 semantic analysis
     for (i = 0; i < modules.dim; i++)
     {
Index: module.c
===================================================================
--- module.c    (revision 336)
+++ module.c    (working copy)
@@ -938,6 +938,8 @@
 {
     size_t len;

+    if(dprogress == 0)
+    return;
     static int nested;
     if (nested)
     return;

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------


More information about the Digitalmars-d-bugs mailing list