1. unable to build OE after the last few commits
- Posted by rkdavis Feb 11, 2015
- 1925 views
I've not been able to build a working OE since 6301:639db32b3548 although i've not finished working my way up the commits to check which one is the start of the problem. I suspect one of Matt Lewis's^H^H^H^H^H^H^H^H^H^Shawn Pringle's last few commits
The build on ARM fails with the following message
me/pi/euphoria/source/euc.ex -nobuild -i /home/pi/euphoria/include -gcc \
-arch ARM -c "/home/pi/euphoria/source/build/eu.cfg" \
-c /home/pi/euphoria/source/eu.cfg /home/pi/euphoria/source/eui.ex )
User supplied library does not exist:
/home/pi/euphoria/source/build/eu.a
/home/pi/euphoria/include/std/mathcons.e:20
<0607>:: The number specified here is too small.
PI = 3.14159_26535_89793_23846,
^
/home/pi/euphoria/source/Makefile.gnu:1100: recipe for target '/home/pi/euphoria /source/build/intobj/main-.c' failed
make[1]: *** [/home/pi/euphoria/source/build/intobj/main-.c] Error 1
make[1]: Leaving directory '/home/pi/euphoria/source'
/home/pi/euphoria/source/Makefile.gnu:546: recipe for target 'interpreter' faile d
make: *** [interpreter] Error 2
Is it just me or is anyone else also failing to build OE on ARM (and/or x86)?
2. Re: unable to build OE after the last few commits
- Posted by ghaberek (admin) Feb 11, 2015
- 1885 views
I successfully built the struct branch the other day. I just switched back to the default branch and I am getting this error, which is different from the error you're getting.
[greg@localhost source]$ hg branch
default
[greg@localhost source]$ make distclean
*SNIP*
[greg@localhost source]$ ./configure
Writing configuration to ./config.gnu...
Getting version string...
VERSION=4.1.0
Writing Makefile...
Writing eu.cfg to /home/greg/Projects/euphoria/source/build/eu.cfg...
[greg@localhost source]$ make
*SNIP*
Translating eui.ex to create
rm -f /home/greg/Projects/euphoria/source/build/intobj/{*.c,*.o}
(cd /home/greg/Projects/euphoria/source/build/intobj;eui -i /home/greg/Projects/euphoria/include /home/greg/Projects/euphoria/source/euc.ex -nobuild -i /home/greg/Projects/euphoria/include -gcc \
-arch ix86_64 -c "/home/greg/Projects/euphoria/source/build/eu.cfg" \
-c /home/greg/Projects/euphoria/source/eu.cfg /home/greg/Projects/euphoria/source/eui.ex )
/home/greg/Projects/euphoria/source/scanner.e:1074
<0074>:: Errors resolving the following references:
'MAX_ATOM' (../../scanner.e:1017) has not been declared.
'MAX_ATOM' (../../scanner.e:1074) has not been declared.
if fnum > (MAX_ATOM - digit)/nBase then
^
Press Enter
-Greg
3. Re: unable to build OE after the last few commits
- Posted by mattlewis (admin) Feb 11, 2015
- 1881 views
I successfully built the struct branch the other day. I just switched back to the default branch and I am getting this error, which is different from the error you're getting.
I'm seeing your error. Looks like Shawn's last commit. He defined MAX_ATOM for 32-bits but not 64. I haven't looked at what he was doing, so I don't know the purpose of any of that.
Matt
4. Re: unable to build OE after the last few commits
- Posted by rkdavis Feb 11, 2015
- 1832 views
just in case it was something hinky on my end i checked out & built the struct branch and then reverted back to default and i still get my original error. what version of eui are you running when building?
5. Re: unable to build OE after the last few commits
- Posted by ghaberek (admin) Feb 11, 2015
- 1851 views
just in case it was something hinky on my end i checked out & built the struct branch and then reverted back to default and i still get my original error. what version of eui are you running when building?
From the struct branch I built last week. Although I originally built that on some struct eubins that were a year or two old. There's probably something ARM-specific that you're running into (an #ifdef or something), just like I'm running into a 64-bit specific issue. I can break out my Raspberry Pi tonight and try building the ARM branch to verify.
-Greg
6. Re: unable to build OE after the last few commits
- Posted by mattlewis (admin) Feb 12, 2015
- 1777 views
Looks like Shawn's last commit. He defined MAX_ATOM for 32-bits but not 64. I haven't looked at what he was doing, so I don't know the purpose of any of that.
Looking at this, there are a lot of red flags here. It looks like we merged 4.0 into the trunk, and lots of 32-bit-isms are in there. I'm not sure we want this merge. Lots of stuff in there, and I'm not sure yet what it was trying to fix with this stuff.
Matt
7. Re: unable to build OE after the last few commits
- Posted by mattlewis (admin) Feb 12, 2015
- 1790 views
I've not been able to build a working OE since 6301:639db32b3548 although i've not finished working my way up the commits to check which one is the start of the problem. I suspect one of Matt Lewis's^H^H^H^H^H^H^H^H^H^Shawn Pringle's last few commits
The build on ARM fails with the following message
me/pi/euphoria/source/euc.ex -nobuild -i /home/pi/euphoria/include -gcc \
-arch ARM -c "/home/pi/euphoria/source/build/eu.cfg" \
-c /home/pi/euphoria/source/eu.cfg /home/pi/euphoria/source/eui.ex )
User supplied library does not exist:
/home/pi/euphoria/source/build/eu.a
/home/pi/euphoria/include/std/mathcons.e:20
<0607>:: The number specified here is too small.
PI = 3.14159_26535_89793_23846,
^
Is it just me or is anyone else also failing to build OE on ARM (and/or x86)?
Looking into this, it looks like Shawn added a lot of stuff for dealing with floating point values. He's testing for underflow, and for some reason, that's happening here, which is obviously incorrect. I suspect that the ARM hardware is finding that because there's more precision in the code than the hardware can handle.
Matt
8. Re: unable to build OE after the last few commits
- Posted by mattlewis (admin) Feb 12, 2015
- 1771 views
Looking into this, it looks like Shawn added a lot of stuff for dealing with floating point values. He's testing for underflow, and for some reason, that's happening here, which is obviously incorrect. I suspect that the ARM hardware is finding that because there's more precision in the code than the hardware can handle.
I've fixed up several errors and merged all the heads together. Most (or maybe all) of Shawn's new tests are failing for me on both 32 and 64 bit Linux builds. Additionally, I'm getting a failure on 64-bit for t_text.e. I assume this is related to tests added for cargoan's patches.
rkdavis, I recommend that you comment out (locally) scanner.e line 1318:
CompileErr(NUMBER_IS_TOO_SMALL)
I think that's a false positive. I'm not really sure why we're testing for this in the scanner, so I think that's pretty harmless. Alternatively, you could probably cross compile and get past this. Though I'm not sure if this same issue won't come up in other places when running on ARM.
Matt
9. Re: unable to build OE after the last few commits
- Posted by rkdavis Feb 12, 2015
- 1783 views
thanks i'll give that a go.
10. Re: unable to build OE after the last few commits
- Posted by mattlewis (admin) Feb 13, 2015
- 1751 views
thanks i'll give that a go.
I've done some more work on the scientific notation stuff and fixed a few tests. The overflow tests are problematic because they're 32-bit specific. I'm not sure if there's an easy way to make the tests more portable (short of just putting in really huge numbers).
Matt
11. Re: unable to build OE after the last few commits
- Posted by rkdavis Feb 16, 2015
- 1716 views
could someone please try
http://www.tybor.com/tflt2int.c
& the ticked answer on http://stackoverflow.com/questions/18644168/catch-floating-point-exceptions-using-a-compiler-option-with-c
on x86 for me please.
