LeafLabs Garden » Topic: Compilation bug?

LeafLabs Garden » Topic: Compilation bug? http://forums.leaflabs.com/topic.php?id=1310 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:27:22 +0000 http://bbpress.org/?v=1.0.2 <![CDATA[Search]]> q http://forums.leaflabs.com/search.php gbulmer on "Compilation bug?" http://forums.leaflabs.com/topic.php?id=1310#post-7971 Tue, 31 Jan 2012 17:38:21 +0000 gbulmer 7971@http://forums.leaflabs.com/ robertm - I really don't know for certain why that macro is there, but Wiring, which was used by the Arduino team, has an integer returning round(). See <a href="http://wiring.org.co/reference/round_.html">Wiring round()</a> Also Java, which influenced Wiring and Processing has an integer returning round See <a href="http://docs.oracle.com/javase/6/docs/api/java/lang/Math.html#round%28double%29">Java round()</a>. I think the Maple/Wiring language is C/C++, so should (as much as practical) leave standard libraries (like math.h) in place so that C/C++ code and documentation works. (full disclosure: I am not a member of LeafLabs staff.) robertm on "Compilation bug?" http://forums.leaflabs.com/topic.php?id=1310#post-7970 Tue, 31 Jan 2012 16:56:26 +0000 robertm 7970@http://forums.leaflabs.com/ gbulmer - Thanks for looking at this. I commented out that line and yes it works correctly afterwords (I'm not sure why the standard definitions are being overridden). I also started messing around with the macro to see if I could get it to work. I changed it to: #define round(x) ((long)(x)>0?((x)+.5):((x)-0.5)) This almost works but I get 123.40 from my function instead of 123.30 I then changed it to: #define round(x) ((long)x>0?((x)+0.5):((x)-0.5)) This works correctly, I get 123.30 from both of my dround functions. I am not sure why changing where the casting takes place and removing the parenthesis from the first x in the ternary should make it work (I think it should be the same)and I don't know how safe removing the parenthesis is, but it does work. It's definitely some weird bug. I'll probably just leave it commented out for now. gbulmer on "Compilation bug?" http://forums.leaflabs.com/topic.php?id=1310#post-7969 Tue, 31 Jan 2012 16:08:44 +0000 gbulmer 7969@http://forums.leaflabs.com/ robertm - I commented out the #define of round(x) in the file wirish_math.h, and your code gives consistent answers. So it is the #define round(x) macro in wirish_math.h which is causing the problem. Unfortunately, someone has chosen to override the standard C definition of round(). So I think it is a bug. gbulmer on "Compilation bug?" http://forums.leaflabs.com/topic.php?id=1310#post-7968 Tue, 31 Jan 2012 14:52:33 +0000 gbulmer 7968@http://forums.leaflabs.com/ I've dug a bit more, and found what I believe is a bug. There is a 'recoding opportunity' (in my Mac version) in the file: /Applications/MapleIDE.app/Contents/Resources/Java/hardware/leaflabs/cores/maple/wirish_math.h: it says <code>#define round(x) ((x)>=0?(long)((x)+0.5):(long)((x)-0.5))</code> which will round a double or float value to a long, thus losing the fractional part. I assume it is the same on all platforms. I tried: <pre><code>#undef round #include <math.h> #undef round</code></pre> but still couldn't get the value correct. Edit: so it might be a combination of a bug and that macro :-( I'd recommend you go find that file, and comment out that 'round' macro. (full disclosure: I am not a member of LeafLabs staff.) gbulmer on "Compilation bug?" http://forums.leaflabs.com/topic.php?id=1310#post-7966 Tue, 31 Jan 2012 13:58:56 +0000 gbulmer 7966@http://forums.leaflabs.com/ robertm - your round functions do not work on my v0.0.12 installation either, so it is not local to your machine. I checked on my Mac, and gcc does generate a program which gives correct, bit identical values. It looks like a compiler bug (or possibly library bug) to me. I modified your code so that I could see the bit patterns, and added a cast to double changing the line: <code>double c = round((num - b) * a) / a;</code> to: <code>double c = round((num - b) * a) / (double) a;</code> and the output matches, and looks correct for this example. <pre><code>#include <math.h> union dbits { double d; uint32 i[2]; }; void printbits(double d) { union dbits a; a.d = d; SerialUSB.print(d); SerialUSB.print(" "); SerialUSB.print(a.i[0], HEX); SerialUSB.println(a.i[1], HEX); } double dround_a(double num, unsigned int ndigits) { int a = pow(10, ndigits); double b = floor(num); // double c = round((num - b) * a) / a; double c = round((num - b) * a) / (double)a; SerialUSB.print("dround_a: c="); printbits(c); double x = b + c; return x; } double dround_b(double num, unsigned int ndigits) { int a = pow(10, ndigits); double b = floor(num); double temp = round((num - b) * a); SerialUSB.print("dround_b: temp="); printbits(temp); double c = temp / a; SerialUSB.print("dround_b: c="); printbits(c); double x = b + c; return x; } void setup(void) { SerialUSB.begin(); } void loop(void) { double value = 123.3456; SerialUSB.println(dround_a(value, 1)); SerialUSB.println(dround_b(value, 1)); delay(1000); }</code></pre> (full disclosure: I am not a member of LeafLabs staff.) robertm on "Compilation bug?" http://forums.leaflabs.com/topic.php?id=1310#post-7964 Tue, 31 Jan 2012 12:55:42 +0000 robertm 7964@http://forums.leaflabs.com/ I've got a function that I am using to round a double to a specific decimal place (example: rounding 123.3456 to 123.3) but isn't working correctly. Below is some test code I wrote. Both functions dround_a() and dround_b() should be exactly the same except dround_b() has the call to round() separated on its own line. When the code is run, dround_a() returns 123.00, while dround_b() returns 123.30 - which is correct. Is this a compiler bug causing dround_a not to work correctly, or is it something else? I'm using the Maple IDE v0.0.12 with a rev.5 Maple board. Here is the test code: <pre><code>#include <math.h> /** * Rounds a number to a specified number of decimal places * * @param num The number to round * @param ndigits The number of decimal places to round to * * @return the rounded value */ double dround_a(double num, unsigned int ndigits) { int a = pow(10, ndigits); double b = floor(num); double c = round((num - b) * a) / a; double x = b + c; return x; } double dround_b(double num, unsigned int ndigits) { int a = pow(10, ndigits); double b = floor(num); double temp = round((num - b) * a); double c = temp / a; double x = b + c; return x; } void setup(void) { } void loop(void) { double value = 123.3456; if (SerialUSB.available()) { uint8 input = SerialUSB.read(); switch(input) { case 'a': SerialUSB.println(dround_a(value, 1)); break; case 'b': SerialUSB.println(dround_b(value, 1)); break; default: break; } SerialUSB.print("> "); } }</code></pre>