[ On , October 21, 2001 at 00:21:01 (-0000), gnats-admin@netbsd.org wrote: ]
> Subject: Re: pkg/14316: strange compiler error while building Mozill-0.9.5 on i386 1.5W
>

Well, I re-installed the comp.tgz set -- the exact same set used on my
P-II which has no problems building mozilla -- and the same error
occured again.

This time I looked at the file and decided that maybe it would be a test
to collapse the wrapped line it was complaining about and then try
continuing the build:

--- work/mozilla/content/html/content/src/nsGenericHTMLElement.cpp.orig	Thu Sep 27 21:38:31 2001
+++ work/mozilla/content/html/content/src/nsGenericHTMLElement.cpp	Sun Oct 21 17:01:14 2001
@@ -1617,9 +1617,7 @@
       result = EnsureWritableAttributes(this, mAttributes, PR_TRUE);
       if (mAttributes) {
         PRInt32   count;
-        result = mAttributes->SetAttributeFor(aAttribute, aValue,
-                                              (NS_STYLE_HINT_CONTENT < impact),
-                                              this, nsnull, count);
+        result = mAttributes->SetAttributeFor(aAttribute, aValue, (NS_STYLE_HINT_CONTENT < impact), this, nsnull, count);
         if (0 == count) {
           ReleaseAttributes(mAttributes);
         }


This time the build continued on until it encountered a supposed syntax
error in another file, line 1058 of mailnews/import/text/src/nsTextAddress.cpp

Sure enough it looked like a real syntax error!  I didn't see that the
last time around!!!  I made the following change/repair:

--- /work/pkgobj/www/mozilla/work/mozilla/mailnews/import/text/src/nsTextAddress.cpp.orig	Sun Oct 21 16:46:11 2001
+++ /work/pkgobj/www/mozilla/work/mozilla/mailnews/import/text/src/nsTextAddress.cpp	Sun Oct 21 16:46:25 2001
@@ -1055,7 +1055,7 @@
 	  break; // 'c'
 
 	case 'd':
-	  if (`-1 != colType.Find("description") )
+	  if ( -1 != colType.Find("description") )
 	  {
 		if (bIsList)
 		  m_database->AddListDescription(newRow, column);

restarted the build, and what do you know if I don't now have a working
mozilla!


I don't understand how or why things happen so differently on to so
similar a pair of machines.  The only difference between them that
should be in any way significant is the processor type (P-Pro vs. P-II).
The former has IDE drives (a Quantum SIROCCO1700A and a Samsung
WA32163A), and the latter has SCSI.  Both machines have ECC DRAM.  The
P-Pro is a clone board, the P-II is an IBM PC Server 325.


Indeed a re-extraction on the P-II machine does not reveal any error as
above in nsTextAddress.cpp.

What could cause a three-bit error (100000 -> 100111) in the extraction,
but apparently only exactly one such error?

(the source .tar.gz was read from NFS in both cases)

What could cause a c++ run fail with a claimed internal error on one
run, but succeed perfectly on another?

Am I still suffering from the basic unreliability of PC hardware, or is
something even more sinister going wrong?


-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>     <woods@robohack.ca>
Planix, Inc. <woods@planix.com>;   Secrets of the Weird <woods@weird.com>