<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent changes to bugs</title><link href="https://sourceforge.net/p/perfinsp/bugs/" rel="alternate"/><link href="https://sourceforge.net/p/perfinsp/bugs/feed.atom" rel="self"/><id>https://sourceforge.net/p/perfinsp/bugs/</id><updated>2012-11-21T15:26:48Z</updated><subtitle>Recent changes to bugs</subtitle><entry><title>bug submitted for dump_trace interface</title><link href="https://sourceforge.net/p/perfinsp/bugs/4/" rel="alternate"/><published>2012-11-21T15:26:48Z</published><updated>2012-11-21T15:26:48Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.netc39365d925f0e268afc7b773ac655f9cf270fb9e</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I am using PI on RHEL6, kernel details : 2.6.32-220.17.1.el6.x86_64&lt;/p&gt;
&lt;p&gt;I meet the compile error on pi_hook.c, as:&lt;br /&gt;
/opt/Software/JProf/Dpiperf/src/driver/pi_hooks.c:1129: error: too many arguments to function dump_trace&lt;/p&gt;
&lt;p&gt;In pi_hoos.c, around 1129 line, it\'s to judge the kernle version to make the appropriate invoke of dump_trace. &lt;br /&gt;
#if (LINUX_VERSION_CODE &amp;gt;= KERNEL_VERSION(2,6,25))&lt;br /&gt;
//bp = frame_pointer(regs);&lt;br /&gt;
dump_trace(NULL, regs, (unsigned long *)stack, bp, &amp;amp;backtrace_ops, (void *)r0s);&lt;br /&gt;
#else&lt;br /&gt;
dump_trace(NULL, regs, (unsigned long *)stack, &amp;amp;backtrace_ops, (void *)r0s);&lt;/p&gt;
&lt;p&gt;It\'s obivious, on this kernel, it invoke the wrong one. This is the interface declare in kernel header, asm/stacktrace.h:&lt;br /&gt;
void dump_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long *stack, const struct stacktrace_ops *ops, void *data);&lt;/p&gt;
&lt;p&gt;Please kindly take a look at this bug.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>spudate bug</title><link href="https://sourceforge.net/p/perfinsp/bugs/3/" rel="alternate"/><published>2009-01-11T09:03:21Z</published><updated>2009-01-11T09:03:21Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.net0a4b11e23916010de1006f6c5c25b78c609bf8c0</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi, &lt;/p&gt;
&lt;p&gt;There is a bug in  spuupdate in checking for the version of symchk. In the file ksyms.c, there's this piece of code: &lt;/p&gt;
&lt;p&gt;xpu_get_file_version(symchk_fqn, &amp;amp;major, &amp;amp;minor, NULL, NULL);&lt;br /&gt;
if (major &amp;gt; 6 || (major == 6 &amp;amp;&amp;amp; minor &amp;gt;= 4)) {&lt;br /&gt;
wrong_file_version_msg(symchk_fqn, major, minor);&lt;br /&gt;
goto TheEnd;&lt;br /&gt;
}&lt;br /&gt;
rc = 1;             // they're both good&lt;/p&gt;
&lt;p&gt;Unfortunately, if the major minor version numbers are as above, then it should be valid for symchk. It shouldn't be calling "wrong_file_version". &lt;/p&gt;
&lt;p&gt;Thanks, &lt;br /&gt;
Arun Vasan &lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>pi_unload.sh can cause system crash</title><link href="https://sourceforge.net/p/perfinsp/bugs/2/" rel="alternate"/><published>2005-11-23T17:06:16Z</published><updated>2005-11-23T17:06:16Z</updated><author><name>Maynard Johnson</name><uri>https://sourceforge.net/u/maynardj/</uri></author><id>https://sourceforge.netd7ec08ff0774b716ab1a00be508ee5fec198087d</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;The following sequence of commands can cause a system&lt;br /&gt;
crash:&lt;br /&gt;
1. pi_load.sh&lt;br /&gt;
2. swtrace init -s 50&lt;br /&gt;
3. swtrace on&lt;br /&gt;
4. swtrace enable 4 17 25 49 50&lt;br /&gt;
5. swtrace it_install -ss&lt;br /&gt;
6. pi_unload.sh&lt;/p&gt;
&lt;p&gt;If you insert 'swtrace off' and 'swtrace it_remove'&lt;br /&gt;
between steps 5 and 6, the pi_unload completes&lt;br /&gt;
successfully.  I will attach a patch to this bug that&lt;br /&gt;
changes the pi_unload.sh script to do a complete&lt;br /&gt;
teardown  in case the user has not done so.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Profiling 64-bit programs on RHEL 4 is broken</title><link href="https://sourceforge.net/p/perfinsp/bugs/1/" rel="alternate"/><published>2005-10-03T22:10:11Z</published><updated>2005-10-03T22:10:11Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.netc020c3e0801ec46ee60755025bd7e0b51d167f01</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Using TPROF on RHEL 4 produces incorrect output when&lt;br /&gt;
profiling 64-bit programs.  The problem can be seen by&lt;br /&gt;
noting incorrect symbols identified within the images&lt;br /&gt;
under test.  The attached file, tprof-2runs.out,&lt;br /&gt;
includes partial reports from two different profiles --&lt;br /&gt;
one for a 32-bit app and one for 64-bit.  The app under&lt;br /&gt;
test, memcpyt, simply does a bunch of memcpy() calls. &lt;br /&gt;
You can see in the Process_Module_Symbol section for&lt;br /&gt;
the 32-bit run that the internal functions&lt;br /&gt;
_wordcopy_fwd_aligned and __GI_memcpy (internal helper&lt;br /&gt;
functions in the callpath of memcpy) have a lot of&lt;br /&gt;
ticks against them under the memcpyt_&amp;lt;pid#&amp;gt;/libc&lt;br /&gt;
subsection.  And for the memcpyt program itself, the&lt;br /&gt;
do_my_memcpy() shows up as hot.  In the 64-bit run, the&lt;br /&gt;
___memset function is incorrectly identified as hot in&lt;br /&gt;
the memcpyt_&amp;lt;pid#&amp;gt;/libc subsection and there are&lt;br /&gt;
"NoSymbols" for memcpyt.&lt;/p&gt;&lt;/div&gt;</summary></entry></feed>