Hey hey all...

I have a problem.
I've downloaded and compiled the boinc client from source.
But now it seems i cant connect to outer world, every request taken will result in a http 413 error (for project attach and acct manager attach)

LOG:
01-May-2008 17:38:50 [http://boinc.bakerlab.org/rosetta/] [work_fetch_debug] best project so far
01-May-2008 17:38:50 [http://boinc.bakerlab.org/rosetta/] [work_fetch_debug] compute_work_requests(): work req 60480.000000, shortfall 60424.499928, urgency Need immediately
01-May-2008 17:38:51 [---] [work_fetch_debug] time_until_work_done(): est 0.000000 ssr 100.000000 apr 1.998165 prs 100.000000
01-May-2008 17:38:51 [http://boinc.bakerlab.org/rosetta/] Sending scheduler request: Requested by user. Requesting 60480 seconds of work, reporting 0 completed tasks
01-May-2008 17:38:51 [---] [http_debug] HTTP_OP::init_post(): http://boinc.bakerlab.org/rosetta_cgi/cgi
01-May-2008 17:38:52 [---] [http_debug] [ID#0] info: About to connect() to boinc.bakerlab.org port 80 (#0)
01-May-2008 17:38:52 [---] [http_debug] [ID#0] info: Trying 140.142.20.103...
01-May-2008 17:38:52 [---] [http_debug] [ID#0] info: Connected to boinc.bakerlab.org (140.142.20.103) port 80 (#0)
01-May-2008 17:38:52 [---] [http_debug] [ID#0] Sent header to server: POST /rosetta_cgi/cgi HTTP/1.1
User-Agent: BOINC client (i486-pc-none 6.1.15)
Host: boinc.bakerlab.org
Accept: */*
Accept-Encoding: deflate, gzip
Content-Type: application/x-www-form-urlencoded
Content-Length: -2147483648

01-May-2008 17:38:52 [---] [http_debug] [ID#0] Received header from server: HTTP/1.1 413 Request Entity Too Large
01-May-2008 17:38:52 [---] [http_debug] [ID#0] Received header from server: Date: Thu, 01 May 2008 15:39:49 GMT
01-May-2008 17:38:52 [---] [http_debug] [ID#0] Received header from server: Server: Apache/2.2.3 (Scientific Linux)
01-May-2008 17:38:52 [---] [http_debug] [ID#0] Received header from server: Connection: close
01-May-2008 17:38:52 [---] [http_debug] [ID#0] Received header from server: Content-Type: text/html; charset=iso-8859-1
01-May-2008 17:38:52 [---] [http_debug] [ID#0] Received header from server:
01-May-2008 17:38:52 [---] [http_xfer_debug] HTTP: wrote 437 bytes
01-May-2008 17:38:52 [---] [http_debug] [ID#0] info: Closing connection #0
01-May-2008 17:38:56 [http://boinc.bakerlab.org/rosetta/] Scheduler request failed: Error 413
01-May-2008 17:38:56 [---] [work_fetch_debug] Request work fetch: RPC complete
01-May-2008 17:38:56 [---] [work_fetch_debug] compute_work_requests(): start
01-May-2008 17:38:56 [---] [work_fetch_debug] compute_work_requests(): cpu_shortfall 60480.000000, overall urgency Need immediately
01-May-2008 17:38:56 [http://boinc.bakerlab.org/rosetta/] [work_fetch_debug] work fetch: project not contactable; skipping


Here my machine:
01-May-2008 17:38:12 [---] Libraries: libcurl/7.18.0 OpenSSL/0.9.8g zlib/1.2.3.3 libidn/1.1
01-May-2008 17:38:12 [---] Data directory: /usr/local/boinc/bin
01-May-2008 17:38:12 [---] Processor: 2 GenuineIntel Intel(R) Pentium(R) III CPU family 1266MHz [Family 6 Model 11 Stepping 1]
01-May-2008 17:38:12 [---] Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse
01-May-2008 17:38:12 [---] OS: Linux: 2.6.24-16-server (Ubuntu 8.04)

My compile options:
./configure --prefix=/usr/local/boinc --disable-server --enable-static --disable-dependency-tracking --enable-bitness=32 CFLAGS='-O3 -fomit-frame-pointer -march=i486 -ffast-math -msse -mfpmath=sse -m32 -malign-double' CXXFLAGS='-O3 -fomit-frame-pointer -march=i486 -ffast-math -msse -mfpmath=sse -m32 -malign-double' --target=i486 --host=i486 --build=i486 --with-libcurl --with-ssl


Does someone know what is wrong?

Content-Length: -2147483648

Woah, what in hell?!

./configure --prefix=/usr/local/boinc --disable-server --enable-static --disable-dependency-tracking --enable-bitness=32 CFLAGS='-O3 -fomit-frame-pointer -march=i486 -ffast-math -msse -mfpmath=sse -m32 -malign-double' CXXFLAGS='-O3 -fomit-frame-pointer -march=i486 -ffast-math -msse -mfpmath=sse -m32 -malign-double' --target=i486 --host=i486 --build=i486 --with-libcurl --with-ssl

Try compiling without optimizations and see if the problem disappears.

---

Try compiling without optimizations and see if the problem disappears.

Yeah, that will work. But that's not what i want.
I recompiled it to gain performance.

I would like to help, if some developer is intrested to squeeze this little bug.
I can also give a shell to the problem computer ;)

Then recompile adding one optimization at a time until you get across the error again. At least then you know which one it is and you can inform the developers.

---

I recompiled it to gain performance.

You won't gain any performance by recompiling the core client. It doesn't compute anything.

---

I recompiled it to gain performance.

You won't gain any performance by recompiling the core client. It doesn't compute anything.

Yes it does.
Because the general client is build as i686, with no optimizations.
I build it as i486, with sse and and math optimalization level 3.

The results:
Stock client:
Number of CPUs: 2
925 floating point MIPS (whetstone) per CPU
1709 integer MIPS (drystone) per CPU

Optimized client:
Number of CPUs: 2
1306 floating point MIPS (whetstone) per CPU
1918 integer MIPS (drystone) per CPU

I recompiled it to gain performance.

You won't gain any performance by recompiling the core client. It doesn't compute anything.

Yes it does.
Because the general client is build as i686, with no optimizations.
I build it as i486, with sse and and math optimalization level 3.

The results:
Stock client:
Number of CPUs: 2
925 floating point MIPS (whetstone) per CPU
1709 integer MIPS (drystone) per CPU

Optimized client:
Number of CPUs: 2
1306 floating point MIPS (whetstone) per CPU
1918 integer MIPS (drystone) per CPU

All the benchmark results tell you is how fast the benchmarks ran :) It definitely can't tell if the science application (completely independent) runs at that speed too. You could very well compile the client with -O3 and the science app with no optimizations at all.

In addition, having higher benchmarks could give you more credits, and pointing fingers calling you a cheater.

---

All the benchmark results tell you is how fast the benchmarks ran :) It definitely can't tell if the science application (completely independent) runs at that speed too. You could very well compile the client with -O3 and the science app with no optimizations at all.

In addition, having higher benchmarks could give you more credits, and pointing fingers calling you a cheater.

True, the client is optimized. Not the work applications. Altrough boinc has to work, optimized or not.

Im not afraid others will call me a cheater. And if they do, i'm pleased to share my custom build.

stat.st_size is type off_t (most likely defined as long)

file_size() casts it to double and assigns it by reference (pointer to) to double

init_post() casts it to integer (which shouldn't be shorter than long on 486)

With save optimisations, this should not destroy the value (even though it's weird to use a double to transport a long into an int)


A reason I could imagine would be different compiler switches for lib and client.