<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Recent changes to bugs</title><link>https://sourceforge.net/p/lmbench/bugs/</link><description>Recent changes to bugs</description><atom:link href="https://sourceforge.net/p/lmbench/bugs/feed.rss" rel="self"/><language>en</language><lastBuildDate>Tue, 07 May 2013 16:23:49 -0000</lastBuildDate><atom:link href="https://sourceforge.net/p/lmbench/bugs/feed.rss" rel="self" type="application/rss+xml"/><item><title>#8 Compiling with "-static" fails trying to link RPC functions</title><link>https://sourceforge.net/p/lmbench/bugs/8/?limit=25#9d85</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;I solved with: LDFLAGS=-ltirpc&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">Zibri</dc:creator><pubDate>Tue, 07 May 2013 16:23:49 -0000</pubDate><guid isPermaLink="false">https://sourceforge.netb0592297ad8bf903fcc4f5f95d97538e276d31ad</guid></item><item><title>Compiling with "-static" fails trying to link RPC functions</title><link>https://sourceforge.net/p/lmbench/bugs/8/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;On some current Linux (Ubuntu and Gentoo) based systems I've tried making the binaries with LDFLAGS=-static and I get errors about missing pmap_* functions (set, unset, etc).  So far the only solution I've had is to remove those from the Makefile as I don't really care about RPC related test results.  Initially I thought adding "-lrpc" might fix it, but apparently that isn't valid or required.  I do see two identical sections in bench.h that are both including rpc/pmap_clnt.h for some reason.  Below is the command and error I see.  Doesn't matter if 32 bit (i686), 64 bit (x64), or arm (armv7l) based system.  Using gcc 4.5.x and 4.7.x on different systems.&lt;/p&gt;
&lt;p&gt;gcc -O -DRUSAGE -DHAVE_uint=1 -DHAVE_int64_t=1 -DHAVE_pmap_clnt_h -DHAVE_socklen_t -DHAVE_DRAND48 -DHAVE_SCHED_SETAFFINITY=1  -static -o ../bin/i686-pc-linux-gnu/lat_select lat_select.c ../bin/i686-pc-linux-gnu/lmbench.a -lm&lt;br /&gt;
lat_select.c: In function ‘server’:&lt;br /&gt;
lat_select.c:134:8: warning: ignoring return value of ‘read’, declared with attribute warn_unused_result [-Wunused-result]&lt;br /&gt;
../bin/i686-pc-linux-gnu/lmbench.a(lib_tcp.o): In function `tcp_connect':&lt;br /&gt;
lib_tcp.c:(.text+0x21b): warning: Using 'gethostbyname' in statically linked applications requires at runtime the shared libraries from the glibc version used for linking&lt;br /&gt;
../bin/i686-pc-linux-gnu/lmbench.a(lib_tcp.o): In function `tcp_done':&lt;br /&gt;
lib_tcp.c:(.text+0x17): undefined reference to `pmap_unset'&lt;br /&gt;
../bin/i686-pc-linux-gnu/lmbench.a(lib_tcp.o): In function `tcp_connect':&lt;br /&gt;
lib_tcp.c:(.text+0x2b1): undefined reference to `pmap_getport'&lt;br /&gt;
../bin/i686-pc-linux-gnu/lmbench.a(lib_tcp.o): In function `tcp_server':&lt;br /&gt;
lib_tcp.c:(.text+0x5b9): undefined reference to `pmap_unset'&lt;br /&gt;
lib_tcp.c:(.text+0x5e2): undefined reference to `pmap_set'&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">Anonymous</dc:creator><pubDate>Tue, 05 Feb 2013 21:56:01 -0000</pubDate><guid isPermaLink="false">https://sourceforge.net1c30e8b8ba7d6cc699a6952697eff0c87d848b65</guid></item><item><title>mount arguments can break summary production</title><link>https://sourceforge.net/p/lmbench/bugs/7/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;Spurious data in 'uninteresting' lines can get accidentally parsed by the getsummary script if they happen to match one of the patterns.&lt;/p&gt;
&lt;p&gt;In my instance, the following line was in our results file:&lt;br /&gt;
[mount: none on /dev type tmpfs (rw,size=64k,mode=755)]&lt;/p&gt;
&lt;p&gt;which I'm guessing matches the rule&lt;br /&gt;
if (/size=64/) {&lt;br /&gt;
in the getsummary script.  This causes a whole slew of data to get ignored for that results file, ending up with only sparse reporting for that run.&lt;br /&gt;
As a quick fix I added the following rule near the top of the getsummary parser loop.&lt;/p&gt;
&lt;p&gt;....&lt;br /&gt;
if (/^\[mount:/) {&lt;br /&gt;
next&lt;br /&gt;
}&lt;br /&gt;
if (/^\[lmbench/) {&lt;br /&gt;
....&lt;/p&gt;
&lt;p&gt;This fixes my particular case, but is pretty horrible and I'm no perl expert, so I'm hoping somebody can come up with a much better perlesque fix.&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">Anonymous</dc:creator><pubDate>Wed, 02 Jul 2008 13:59:18 -0000</pubDate><guid isPermaLink="false">https://sourceforge.net16c031fafe561af18a9f759f9c5cd5ad901df72a</guid></item><item><title>lmdd Fails on Fails to output files over 2 GB</title><link>https://sourceforge.net/p/lmbench/bugs/6/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;When you run lmdd outputting a file over 2 GB, the lmdd process will fail.  Here is the output of a failed lmdd:&lt;/p&gt;
&lt;p&gt;lmdd if=/dev/zero of=/mnt/tmp/metadata.bin bs=4m move=4g&lt;br /&gt;
write: wanted=4194304 got=1282048&lt;br /&gt;
2080.3748 MB in 42.9432 secs, 48.4448 MB/sec&lt;/p&gt;
&lt;p&gt;To fix this, you need to add these 3 defines to the makefile:&lt;br /&gt;
-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64&lt;/p&gt;
&lt;p&gt;In lmdd.c at the getfile method where you are creating the output flags, you need to add the following lines of code:&lt;/p&gt;
&lt;p&gt;#if defined(O_LARGEFILE)&lt;br /&gt;
// Linux requires this flag to open a file larger than 2^31 bytes&lt;br /&gt;
oflags |= O_LARGEFILE ;&lt;br /&gt;
#endif&lt;/p&gt;
&lt;p&gt;This bug may affect other lmbench programs, but I haven't tried them.&lt;/p&gt;
&lt;p&gt;Thanks&lt;/p&gt;
&lt;p&gt;jesse.anderson*REMOVETHIS*@4bright.com&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">Anonymous</dc:creator><pubDate>Wed, 19 Dec 2007 17:47:16 -0000</pubDate><guid isPermaLink="false">https://sourceforge.net34190c8b77a803c8176b586d203aee8b61a2457f</guid></item><item><title>lat_mem_rd:step() infinite loop</title><link>https://sourceforge.net/p/lmbench/bugs/5/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;If k is 2048 or greater in step, then step will be caught in an infinite loop in the following line of code.&lt;/p&gt;
&lt;p&gt;for (s = 32 * 1024; s &amp;lt;= k; s *= 2)&lt;br /&gt;
;&lt;/p&gt;
&lt;p&gt;k = 0x80000000 and eventually s will equal 0x80000000. The next multiplication (shl) will set s equal to zero. &lt;/p&gt;
&lt;p&gt;Proposed fix: prevent k from being set to 2048 or greater . There are two places this should be done. First, the CONFIG file should set MB to a value less that 2048. Second, lat_mem_rd should reset any value greater than or equal to 2048 to 2047. &lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">rmlane</dc:creator><pubDate>Wed, 08 Aug 2007 17:27:41 -0000</pubDate><guid isPermaLink="false">https://sourceforge.netfe70b667ae3211f6c5a77c96e415fbd7a5a1739f</guid></item><item><title>No error checking</title><link>https://sourceforge.net/p/lmbench/bugs/4/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;Hi,&lt;/p&gt;
&lt;p&gt;I am getting a hang with lmbench 3.0, latest release, when using lat_mem_rd -N 1 96m 256&lt;br /&gt;
&lt;/p&gt;
&lt;p&gt;After looking at the code, I noticed that lib_timing.c makes several calls to read() &amp;amp; write() but never checks the return code for success or number of bytes read/written. I think this could be my problem...since the return codes are never checked, if any of these&lt;br /&gt;
calls fail, it is never retried....?  Which hangs the parent waiting for the child(s), or hangs the child(s) waiting for the parent.&lt;/p&gt;
&lt;p&gt;Any reason why the return codes are not checked?  Obviously, read()/write() can fail or only process a portion of the bytes.&lt;/p&gt;
&lt;p&gt;Dave&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">Anonymous</dc:creator><pubDate>Wed, 04 Apr 2007 15:29:24 -0000</pubDate><guid isPermaLink="false">https://sourceforge.netda46076d52ffef1628270f86e5c306626e1a4c96</guid></item><item><title>Lmbench scheduler increases latency</title><link>https://sourceforge.net/p/lmbench/bugs/3/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;Lmbench scheduler in lmbench-3.0-a7 is useful to get stable, reproducible results in SMP. But it was detected (see &lt;a href="http://marc.theaimsgroup.com/?l=linux-kernel&amp;amp;m=116304920822136&amp;amp;w=2\" rel="nofollow"&gt;http://marc.theaimsgroup.com/?l=linux-kernel&amp;amp;m=116304920822136&amp;amp;w=2\&lt;/a&gt;) that fork() latency reported by lmbench is much higher than real latency of fork()+sched_setaffinity(). It is because of a lot of unneeded functions is called in interval of time which was measured as a fork() function latency. The functions which could be moved out of time interval been measured are: getenv(), sysconf(), strcasecmp() (7 calls), malloc(), realloc() (in a loop) and sched_getaffinity(). &lt;br /&gt;
Lmbench is able to measure latency for Linux, Solaris, HPUX, IRIX, AIX but for Linux 35 lines are added in interval been measured including calls to malloc() twice, sched_getaffinity() and realloc() calls in a loop. As a result lmbench reports latency value 410 microseconds while correct latency 138 microseconds.&lt;br /&gt;
The patch for issue foxing is attached.&lt;br /&gt;
Leonid&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">Anonymous</dc:creator><pubDate>Mon, 27 Nov 2006 14:55:18 -0000</pubDate><guid isPermaLink="false">https://sourceforge.net30a1a29d5c528c8c34d8f4d63051269cb11fb54b</guid></item><item><title>Lmbenchmarking:  Tests hang while running ./config-run</title><link>https://sourceforge.net/p/lmbench/bugs/2/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;Here I was trying to do lmbenchmarking on a machine &lt;br /&gt;
with more than 3 GB ram.&lt;/p&gt;
&lt;p&gt;Steps to reproduce :-&lt;/p&gt;
&lt;p&gt;1. cd to /opt/lmbench/scripts&lt;/p&gt;
&lt;p&gt;2. run ./config-run.&lt;/p&gt;
&lt;p&gt;./config-run hangs on doing a memory check.&lt;/p&gt;
&lt;p&gt;More info :-&lt;/p&gt;
&lt;p&gt;Memtest works fine with values up to 2047MB. After &lt;br /&gt;
that, memsize hangs in an infinite loop:&lt;/p&gt;
&lt;p&gt;The fault lies in the function `timeit' in &lt;br /&gt;
`memsize.c' in the following loop:&lt;/p&gt;
&lt;p&gt;for (s = 8 * 1024 * 1024; s &amp;lt;= range; s *= 2)&lt;br /&gt;
;&lt;/p&gt;
&lt;p&gt;`s' is declared of type size_t and eventually is &lt;br /&gt;
reduced to zero here, causing an infinite loop. One &lt;br /&gt;
solution is to declare `s' as `unsigned long long' &lt;br /&gt;
instead of `size_t'. Now memsize successfully checks &lt;br /&gt;
memory past 2048 MB. I saw another bug where it will &lt;br /&gt;
not check memory past 2730 MB.&lt;/p&gt;
&lt;p&gt;Thanks and regards,&lt;br /&gt;
San&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">san</dc:creator><pubDate>Thu, 17 Aug 2006 07:54:45 -0000</pubDate><guid isPermaLink="false">https://sourceforge.netfb874783d4e6be8bdd89e071bf79c93f86b65d8d</guid></item><item><title>lmbench-3.0-a3: bugs in scripts/config-run</title><link>https://sourceforge.net/p/lmbench/bugs/1/</link><description>&lt;div class="markdown_content"&gt;&lt;p&gt;There are two bugs I found when executing&lt;br /&gt;
make results&lt;/p&gt;
&lt;p&gt;The first one is in line 115 of scripts/config-run:&lt;br /&gt;
MB = $TOTAL_MEM&lt;br /&gt;
This causes an error on my SLES8 installation (zLinux).&lt;br /&gt;
It works after removing the blanks:&lt;br /&gt;
MB=$TOTAL_MEM&lt;br /&gt;
I think other Linux distributions might work with the&lt;br /&gt;
blanks left in, but the version without blanks should&lt;br /&gt;
work everywhere.&lt;/p&gt;
&lt;p&gt;The second one is in lines 128ff of scripts/config-run:&lt;/p&gt;
&lt;p&gt;echo $ECHON "MB [default $MB] $ECHOC"&lt;br /&gt;
read TMP&lt;br /&gt;
if [ X$TMP != X ]&lt;br /&gt;
then    TOTAL_MEM=$TMP&lt;br /&gt;
fi&lt;br /&gt;
# Certain machines tend to barf when you try and bcopy 8MB.&lt;br /&gt;
# Figure out how much we can use.&lt;br /&gt;
echo "Checking to see if you have $MB MB; please wait&lt;br /&gt;
for a moment..."&lt;br /&gt;
AVAILMEM=`../bin/$OS/memsize $MB`&lt;br /&gt;
AVAILMEM=`../bin/$OS/memsize $MB`&lt;br /&gt;
AVAILMEM=`../bin/$OS/memsize $MB`&lt;/p&gt;
&lt;p&gt;TOTAL_MEM is set to the value the user enters, but the&lt;br /&gt;
memory check is performed against the $MB value that&lt;br /&gt;
the system generates, so the user override never works.&lt;br /&gt;
Older versions of LMBench contain the following lines:&lt;/p&gt;
&lt;p&gt;if [ X$TMP != X ]&lt;br /&gt;
then    MB=$TMP&lt;br /&gt;
fi&lt;/p&gt;
&lt;p&gt;so probably both MB and TOTAL_MEM should be set.&lt;/p&gt;&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">tstaudt</dc:creator><pubDate>Wed, 31 Mar 2004 11:44:41 -0000</pubDate><guid isPermaLink="false">https://sourceforge.netddad5fc6d7e8ffe8aba4db41af2d510dfef12d8e</guid></item></channel></rss>