Post by thread:[PIC]: Bug in gpasm's expression evaluator?

[CC'd to the GPUtils project admins] Hi, I think I may have found a bug in GPAsm that relates to its decoding of hexadecimal expressions. Specifically, if I have a constant like this: DELTA EQU 0x9E3779B9 .. GPasm seems to clip it at 0x7FFFFFFF. So i do this with the constant: MOVLW (((DELTA & 0xFF000000)>>8)>>8)>>8 .. I end up getting the value 0x7F loaded into W. MPASM, on the other hand, evaluates the expression correctly and loads the value 0x9E into W. This particular bug is wreaking absolute havoc with my 32-bit math macros - especially with the "Load 32-bit constant into variable" (M_COPYLIT) macro. Does a workaround for this bug exist - besides, of course, "Rip out the macros and do the calculations manually"? I suppose I could use MPASM under Dosemu or WINE, but IME Dosemu takes ages to load. I can provide a full, compilable example to demonstrate this bug. Also, if I try to do a full shift-right-by-24-bits (i.e. "DELTA >> 24"), GPAsm rejects the line of code with an error - something about the 24 being too large? Thanks. -- Phil. | Acorn Risc PC600 Mk3, SA202, 64MB, 6GB, spam_OUTphilpemTakeThisOuTdsl.pipex.com | ViewFinder, Ethernet (Acorn AEH62), http://www.philpem.dsl.pipex.com/ | 8xCD, framegrabber, Teletext Nothing can go wroff001010 bus error: core dumped -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

----- Original Message ----- From: Philip Pemberton <.....philpemKILLspam@spam@DSL.PIPEX.COM> To: <PICLISTKILLspamMITVMA.MIT.EDU> Sent: Monday, October 20, 2003 4:10 PM Subject: [PIC]: Bug in gpasm's expression evaluator? {Quote hidden}> [CC'd to the GPUtils project admins] > > Hi, > I think I may have found a bug in GPAsm that relates to its decoding of > hexadecimal expressions. Specifically, if I have a constant like this: > DELTA EQU 0x9E3779B9 > .. GPasm seems to clip it at 0x7FFFFFFF. So i do this with the constant: > MOVLW (((DELTA & 0xFF000000)>>8)>>8)>>8 > .. I end up getting the value 0x7F loaded into W. MPASM, on the other hand, > evaluates the expression correctly and loads the value 0x9E into W. This > particular bug is wreaking absolute havoc with my 32-bit math macros - > especially with the "Load 32-bit constant into variable" (M_COPYLIT) macro. > Does a workaround for this bug exist - besides, of course, "Rip out the > macros and do the calculations manually"? I suppose I could use MPASM under > Dosemu or WINE, but IME Dosemu takes ages to load. > I can provide a full, compilable example to demonstrate this bug. Also, if > I try to do a full shift-right-by-24-bits (i.e. "DELTA >> 24"), GPAsm rejects > the line of code with an error - something about the 24 being too large? > > Thanks. > -- > Phil. | Acorn Risc PC600 Mk3, SA202, 64MB, 6GB, > .....philpemKILLspam.....dsl.pipex.com | ViewFinder, Ethernet (Acorn AEH62), > http://www.philpem.dsl.pipex.com/ | 8xCD, framegrabber, Teletext > Nothing can go wroff001010 bus error: core dumped Hi Phil, I've seen this problem before (not in GPASM BTW). It is to do with using strtol to convert a string to a long. Hunt down strtol in the GPASM source code and replace them with strtoul (string to unsigned long) this should fix your problem. Regards Sergio Masci http://www.xcprod.com/titan/XCSB - optimising structured PIC BASIC compiler -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

In message <EraseMEPine.LNX.4.44.0310200903320.12579-100000spam_OUTTakeThisOuTruckus.brouhaha.com>> Scott Dattalo <scottspam_OUTDATTALO.COM> wrote: > Is this in the version 0.11.7 that Craig released yesterday? It was when I last checked. It also seems to be in the current CVS version. Thanks. -- Phil. | Acorn Risc PC600 Mk3, SA202, 64MB, 6GB, @spam@philpemKILLspamdsl.pipex.com | ViewFinder, Ethernet (Acorn AEH62), http://www.philpem.dsl.pipex.com/ | 8xCD, framegrabber, Teletext Ever noticed how DOS never says "GOOD command or filename"? -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

In message <1066696096.1419.9.camel@R2D2> Craig Franklin <KILLspamcraigfranklinKILLspamcomcast.net> wrote: > I believe this bug is covered by: > > https://sourceforge.net/tracker/index.php?func=detail&aid=716633&group_id=41924&atid=431665 > > I will take care of if. Give me a little time. It looks like the glitch can be fixed by replacing the strtol in "gpasm/util.c" with a strtoul. At least, that's what killed the bug in my copy of GPASM. Thank Sergio Masci, not me :) Anyway, thanks to Scott Dattalo and Craig Franklin (and anyone else who contributed code to GPUtils) - GPUtils is great! :) Later. -- Phil. | Acorn Risc PC600 Mk3, SA202, 64MB, 6GB, RemoveMEphilpemTakeThisOuTdsl.pipex.com | ViewFinder, Ethernet (Acorn AEH62), http://www.philpem.dsl.pipex.com/ | 8xCD, framegrabber, Teletext For sale: One Russian space station. Needs minor repairs. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

pic microcontroller discussion list wrote: >> I will take care of if. Give me a little time. > It looks like the glitch can be fixed by replacing the strtol in > "gpasm/util.c" with a strtoul. At least, that's what killed the bug > in my copy of GPASM. Thank Sergio Masci, not me :) And I hope someone ran a stopwatch on that bug report, debug, repair cycle- viva Open Source :-) :-) Nigel -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

On Tue, 2003-10-21 at 01:38, Philip Pemberton wrote: > In message <1066696096.1419.9.camel@R2D2> > Craig Franklin <spamBeGonecraigfranklinspamBeGonecomcast.net> wrote: > > > I believe this bug is covered by: > > > > sourceforge.net/tracker/index.php?func=detail&aid=716633&group_id=41924&atid=431665 > > > > I will take care of if. Give me a little time. > It looks like the glitch can be fixed by replacing the strtol in > "gpasm/util.c" with a strtoul. At least, that's what killed the bug in my > copy of GPASM. Thank Sergio Masci, not me :) > Anyway, thanks to Scott Dattalo and Craig Franklin (and anyone else who > contributed code to GPUtils) - GPUtils is great! :) > A patch has been committed to CVS. It will be in the next gputils release. That release is a couple weeks away. If you can't wait, the change will be part of the CVS snapshot tomorrow. A link to the snapshot is located at http://gputils.sourceforge.net. You also had a problem with ">> 24". I think it is caused by using a default radix of hex. gpasm is reading this as ">> 0x24" or ">> 36". Change your radix to dec or use d'24'. {Quote hidden}> Later. > -- > Phil. | Acorn Risc PC600 Mk3, SA202, 64MB, 6GB, > TakeThisOuTphilpemEraseMEspam_OUTdsl.pipex.com | ViewFinder, Ethernet (Acorn AEH62), > http://www.philpem.dsl.pipex.com/ | 8xCD, framegrabber, Teletext > For sale: One Russian space station. Needs minor repairs. > > -- > http://www.piclist.com hint: The PICList is archived three different > ways. See http://www.piclist.com/#archives for details. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.