#3904: Bug in shipped gcc makes profiling lose resolution
-----------------------------+----------------------------------------------
Reporter: augustss | Owner:
Type: bug | Status: new
Priority: normal | Milestone:
Component: Build System | Version: 6.12.1
Keywords: | Difficulty:
Os: Windows | Testcase:
Architecture: x86 | Failure: None/Unknown
-----------------------------+----------------------------------------------
Comment(by simonmar):
Replying to [comment:1 augustss]:
> Why is the prining so complex anyway, why not just
> fprintf(..., "%s %.02f\n", ..., sampleValue);
> ?
This is the patch that did it:
{{{
Tue May 11 19:36:10 BST 2004 panne
* [project @ 2004-05-11 18:36:10 by panne]
Make the printing of samples really locale-independent
{
hunk ./ghc/rts/ProfHeap.c 2
- * $Id: ProfHeap.c,v 1.51 2004/03/19 23:20:20 panne Exp $
+ * $Id: ProfHeap.c,v 1.52 2004/05/11 18:36:10 panne Exp $
hunk ./ghc/rts/ProfHeap.c 35
+#include <math.h>
hunk ./ghc/rts/ProfHeap.c 386
+static void
+printSample(rtsBool beginSample, StgDouble sampleValue)
+{
+ StgDouble fractionalPart, integralPart;
+ fractionalPart = modf(sampleValue, &integralPart);
+ fprintf(hp_file, "%s %d.%02d\n",
+ (beginSample ? "BEGIN_SAMPLE" : "END_SAMPLE"),
+ (int)integralPart, (int)(fractionalPart * 100 + 0.5));
+}
+
hunk ./ghc/rts/ProfHeap.c 457
- fprintf(hp_file, "BEGIN_SAMPLE 0.00\n");
- fprintf(hp_file, "END_SAMPLE 0.00\n");
+ printSample(rtsTrue, 0);
+ printSample(rtsFalse, 0);
hunk ./ghc/rts/ProfHeap.c 500
- fprintf(hp_file, "BEGIN_SAMPLE %0.2f\n", seconds);
- fprintf(hp_file, "END_SAMPLE %0.2f\n", seconds);
+ printSample(rtsTrue, seconds);
+ printSample(rtsFalse, seconds);
hunk ./ghc/rts/ProfHeap.c 597
+ (void)p; /* keep gcc -Wall happy */
hunk ./ghc/rts/ProfHeap.c 744
- fprintf(hp_file, "BEGIN_SAMPLE %0.2f\n", census->time);
+ printSample(rtsTrue, census->time);
hunk ./ghc/rts/ProfHeap.c 757
- fprintf(hp_file, "END_SAMPLE %0.2f\n", census->time);
+ printSample(rtsFalse, census->time);
hunk ./ghc/rts/ProfHeap.c 837
- fprintf(hp_file, "END_SAMPLE %0.2f\n", census->time);
+ printSample(rtsFalse, census->time);
}
}}}
I guess the problem is that the decimal point is not necessarily a '.'.
It looks like a later patch changed the types to StgWord64, which tickled
the bug in gcc.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/3904#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
