ben@epmooch.UUCP (Rev. Ben A. Mesander) (11/17/90)
>In article <146@screamer.csee.usf.edu> stelmack@screamer.csee.usf.edu (Gregory M. Stelmack) writes: >In article <1990Nov16.201320.4842@servalan.uucp> ben@servalan.uucp (Ben Mesander) writes: >>__chip is impossible with GNU CC (mybe with atom, though?). However, >>I bought Lattice and now the SAS upgrade. I have yet to find a non-trivial >>program that the optimizer works on, I constantly hit compiler limits and > >Please define non-trivial. The global optimizer worked fine on my Spades >program and has yet to have a problem with all the database code I've been >working on. > Examples of non-trivial programs that are broken when you use the SAS optimizer: MicroEMACS egrep sed awk teco GNU chess pdtar >-- Greg Stelmack -- | ben@epmooch.UUCP (Ben Mesander) | "Cash is more important than | | ben%servalan.UUCP@uokmax.ecn.uoknor.edu | your mother." - Al Shugart, | | !chinet!uokmax!servalan!epmooch!ben | CEO, Seagate Technologies |
stelmack@screamer.csee.usf.edu (Gregory M. Stelmack) (11/17/90)
In article <1990Nov16.201320.4842@servalan.uucp> ben@servalan.uucp (Ben Mesander) writes: >__chip is impossible with GNU CC (mybe with atom, though?). However, >I bought Lattice and now the SAS upgrade. I have yet to find a non-trivial >program that the optimizer works on, I constantly hit compiler limits and Please define non-trivial. The global optimizer worked fine on my Spades program and has yet to have a problem with all the database code I've been working on. >bugs, I have had it spit up mysterious internal error messages that are not >in the manual, I have repeatedly found bugs in the library routines, the >code generation is only fair, and the library is inadequate for my needs. >In short, if I called SAS whenever I had a problem with thier compiler or >library, I'd be broke. The only time I ever hit a compiler limitation was when I tried to compile the ray2 program from DBWRender -- nothing else. For Amiga development, I find the library quite adequate. The only time I have problems with the library is when going between Amiga and UNIX -- since GCC is a UNIX compiler, this is the situation where GCC might prove invaluable. But, for Amiga stuff (and basic UNIX programs), SAS works quite well for me. I would like to see more UNIX extensions in the SAS library (but notice that they added several this time around, and I suspect we'll see more with each upgrade). > >ben@epmooch.UUCP Just presenting an opposing viewpoint, -- Greg Stelmack -- Email: stelmack@sol.csee.usf.edu -- USmail: USF Box 1510, Tampa, FL 33620-1510 -- Amiga: the only way to compute! -- IRAQNOPHOBIA: Nothing a little RAID wouldn't cure!
mab@drutx.ATT.COM (Alan Bland) (11/18/90)
In article <1990Nov16.201320.4842@servalan.uucp> ben@servalan.uucp (Ben Mesander) writes: >I bought Lattice and now the SAS upgrade. I have yet to find a non-trivial >program that the optimizer works on, I constantly hit compiler limits and I've run the SAS 5.10 global optimizer on a 20,000 line (non-comments) C program that does graphics, MIDI I/O, hardware timer interrupts, and other goodies, and the results are just peachy. I'd call that program non-trivial. Of course, I did hit the #pragma bug (odd constants passed to system calls with lots of args will crash the compiler), but once I got beyond that, no problems. Ever since Lattice introduced the global optimizer, it has produced code that works the same as without the optimizer, and the code is smaller and presumably faster (though I haven't really noticed a speed difference in the programs I've tried). -- -- Alan Bland -- att!drutx!mab == mab@drutx.ATT.COM -- AT&T Bell Laboratories, Denver CO -- (303)538-3510
ben@epmooch.UUCP (Rev. Ben A. Mesander) (11/19/90)
>In article <15941@cbmvax.commodore.com> jesup@cbmvax.commodore.com (Randell Jesup) writes: >In article <ben.3496@epmooch.UUCP> ben@epmooch.UUCP (Rev. Ben A. Mesander) writes: >>Examples of non-trivial programs that are broken when you use the SAS >>optimizer: >> >>MicroEMACS >>egrep >>sed >>awk >>teco >>GNU chess >>pdtar > > Have you bothered figuring out why those programs broke? Most things >I've seen 'broken' by the optimizer was code that was already buggy but >happened to work because the compiler wasn't as agressive, or because of >side-effects. There may be optimizer bugs, but if there are you should >report them (SAS is very good about fixing bugs if given an example). Fine. I don't have the time to deal with it (I work full time, and I'm a graduate student). Yes, I've looked at the 68000 code with CPR, and it appears that I can spoof SAS C into generating incorrect offsets for automatic variables with the optimizor on. I have not noted a particular pattern of events that lead to this error, other than the optimizor on: code is incorrect, optimizor off: code is correct. >Personally, I always use the optimizer, and have yet to hit a bug in it >personally. There were some lc2 bugs (in some betas) that only showed up >if the optimizer was on - perhaps this is some of what you're seeing. Personally, I'm very happy it works for you :-). I'm sick to death of people saying "It hasn't happened to me, so it must not exist!" It does have bugs. > Lastly, which version are you using (when you last tried those)? 5.10 5.10 5.05 5.10 5.10 5.10 5.05 >Randell Jesup, Keeper of AmigaDos, Commodore Engineering. -- | ben@epmooch.UUCP (Ben Mesander) | | ben%servalan.UUCP@uokmax.ecn.uoknor.edu | | !chinet!uokmax!servalan!epmooch!ben | SAS bug of the month club: Open a file that already exists. Now try an access(file,0) call to see if it exists. It doesn't exist? No? Go look in the directory. Sure 'nuff, it's there. Looks to me like someone said: "Lock() failed? Doesn't exist!" without checking the return value to see if the "object was in use".
jesup@cbmvax.commodore.com (Randell Jesup) (11/19/90)
In article <ben.3496@epmooch.UUCP> ben@epmooch.UUCP (Rev. Ben A. Mesander) writes: >Examples of non-trivial programs that are broken when you use the SAS >optimizer: > >MicroEMACS >egrep >sed >awk >teco >GNU chess >pdtar Have you bothered figuring out why those programs broke? Most things I've seen 'broken' by the optimizer was code that was already buggy but happened to work because the compiler wasn't as agressive, or because of side-effects. There may be optimizer bugs, but if there are you should report them (SAS is very good about fixing bugs if given an example). Personally, I always use the optimizer, and have yet to hit a bug in it personally. There were some lc2 bugs (in some betas) that only showed up if the optimizer was on - perhaps this is some of what you're seeing. Lastly, which version are you using (when you last tried those)? -- Randell Jesup, Keeper of AmigaDos, Commodore Engineering. {uunet|rutgers}!cbmvax!jesup, jesup@cbmvax.commodore.com BIX: rjesup Thus spake the Master Ninjei: "If your application does not run correctly, do not blame the operating system." (From "The Zen of Programming") ;-)
ben@epmooch.UUCP (Rev. Ben A. Mesander) (11/20/90)
>In article <2024@shodha.enet.dec.com> ridder@elvira.enet.dec.com (Hans Ridder) writes: [much discussion of SAS/C] >Wow! I don't think Randell was suggesting that the problem you were >having didn't exist, just that it might have been fixed. You say you >don't have time to deal with it, but you apparently have time to let us >all know how much trouble everything is for you. Randell was trying to >help you with your problem. It appears that you really don't want help, >you just want to complain. You're right, I will call SAS. I have a non-beta SAS/C 5.10. I have been experiencing optimizor problems with the compiler since the first program I have compiled with Lattice C 5.05 (MicroEMACS 3.10). I like your tone - you must be real good for DEC customer support (1/2 :-) Reading news takes very little effort on my part; I can do it late at night. Calling SAS requires that I take time out of my day as a busy Computer Engineer and part-time graduate student and make a long-distance phone call. Sometimes, reporting a compiler bug for something that is my personal "toy" doesn't take precedence over keeping a running computer shop for 60 users. >I'll bet the whole net is "sick to death" of people saying this or that >is junk or broken, when they won't take appropriate steps to take care >of it, except to tell everyone on the net how clever they were to find a >bug. I've never thought of myself as "clever". I'd rather not find a bug in my compiler, thank you. If I thought SAS C was "junk", I wouldn't have invested a large chunk of money in it. I *do* think that SAS has a long way to go until they have a relatively bug-free compiler and library. That is my _personal opinion_. > Hans-Gabriel Ridder Digital Equipment Corporation -- | ben@epmooch.UUCP (Ben Mesander) | "Cash is more important than | | ben%servalan.UUCP@uokmax.ecn.uoknor.edu | your mother." - Al Shugart, | | !chinet!uokmax!servalan!epmooch!ben | CEO, Seagate Technologies |
ken@cbmvax.commodore.com (Ken Farinsky - CATS) (11/21/90)
In article <ben.3508@epmooch.UUCP> ben@epmooch.UUCP (Rev. Ben A. Mesander) writes: > >SAS bug of the month club: >Open a file that already exists. Now try an access(file,0) call to see >if it exists. It doesn't exist? No? Go look in the directory. Sure 'nuff, >it's there. Looks to me like someone said: "Lock() failed? Doesn't exist!" >without checking the return value to see if the "object was in use". Have you considered sending bug reports (along with sample code) to SAS? Or are you just going to complain that it is broken? Developers can only fix problems that: 1) they know about, and 2) they can reproduce. -- -- Ken Farinsky - CATS - (215) 431-9421 - Commodore Business Machines uucp: ken@cbmvax.commodore.com or ...{uunet,rutgers}!cbmvax!ken bix: kfarinsky
ridder@elvira.enet.dec.com (Hans Ridder) (11/21/90)
In article <ben.3508@epmooch.UUCP> ben@epmooch.UUCP (Rev. Ben A. Mesander) writes: >>In article <15941@cbmvax.commodore.com> jesup@cbmvax.commodore.com (Randell Jesup) writes: >> Have you bothered figuring out why those programs broke? Most >>things I've seen 'broken' by the optimizer was code that was already >>buggy but happened to work because the compiler wasn't as agressive, or >>because of side-effects. There may be optimizer bugs, but if there are >>you should report them (SAS is very good about fixing bugs if given an >>example). > >Fine. I don't have the time to deal with it (I work full time, and I'm >a graduate student). Yes, I've looked at the 68000 code with CPR, and >it appears that I can spoof SAS C into generating incorrect offsets for >automatic variables with the optimizor on. I have not noted a particular >pattern of events that lead to this error, other than the optimizor on: >code is incorrect, optimizor off: code is correct. > >>Personally, I always use the optimizer, and have yet to hit a bug in it >>personally. There were some lc2 bugs (in some betas) that only showed up >>if the optimizer was on - perhaps this is some of what you're seeing. > >Personally, I'm very happy it works for you :-). I'm sick to death of >people saying "It hasn't happened to me, so it must not exist!" It does >have bugs. Wow! I don't think Randell was suggesting that the problem you were having didn't exist, just that it might have been fixed. You say you don't have time to deal with it, but you apparently have time to let us all know how much trouble everything is for you. Randell was trying to help you with your problem. It appears that you really don't want help, you just want to complain. I'll bet the whole net is "sick to death" of people saying this or that is junk or broken, when they won't take appropriate steps to take care of it, except to tell everyone on the net how clever they were to find a bug. The point I think you're missing in all of this is that it doesn't do anyone any good for you to tell us that SAS/C is broken, or a lousy compiler, or even that GCC is better, if you're not willing to report the problems/bugs to the proper people. Usenet isn't SAS/C support. >| ben@epmooch.UUCP (Ben Mesander) | >SAS bug of the month club: >Open a file that already exists. Now try an access(file,0) call to see >if it exists. It doesn't exist? No? Go look in the directory. Sure 'nuff, >it's there. Looks to me like someone said: "Lock() failed? Doesn't exist!" >without checking the return value to see if the "object was in use". Have you reported it? (*Sigh*) -hans ------------------------------------------------------------------------ Hans-Gabriel Ridder Digital Equipment Corporation ridder@elvira.enet.dec.com Customer Support Center ...decwrl!elvira.enet!ridder Colorado Springs, CO
didierj@swindj.UUCP (Alain Didierjean) (11/21/90)
In article <15983@cbmvax.commodore.com>, Ken Farinsky - CATS writes: -> In article <ben.3508@epmooch.UUCP> ben@epmooch.UUCP (Rev. Ben A. Mesander) writes: -> > -> >SAS bug of the month club: -> >[text deleted here....] -> Have you considered sending bug reports (along with sample code) to SAS? -> Or are you just going to complain that it is broken? Developers can -> only fix problems that: 1) they know about, and 2) they can reproduce. -> -- Which brings a question already asked to John Toebes a month ago without any answer to-day (hope he got the mail). YES there are some bugs in SAS C 5.10, a lot of them, usually minor. YES there are people here (Europe) eager to get the compiler they rely on improved quickly. A lot of them being members of Amiga Developpers Support Program in Europe (ADSPE). Considering that the only practical and not too expensive way to get in touch with SAS people from here is this very net, could we possibly get an E-Mail address for SAS Institute Amiga C debugging crew... Or any organisation that agree to be used as a relay of some sort... This could possibly be useful to some american users too. If not, I'm going to join the "SAS bug of the month club". Not very efficient, but excellent for mental hygien.. -> Ken Farinsky - CATS - (215) 431-9421 - Commodore Business Machines -> uucp: ken@cbmvax.commodore.com or ...{uunet,rutgers}!cbmvax!ken -> bix: kfarinsky ______________________________________________________________ // \ // Alain DIDIERJEAN The Software Winos won't die \ // The Software Winery Their Spirits will be there forever \ //___________________________________________________________________\ // \ // ...cbmvax!cbmehq!cbmfra!swindj!didierj \ ==========================================================================
micke@slaka.sirius.se (Mikael Karlsson) (11/26/90)
In article <183e09fb.ARN03933@swindj.UUCP>, didierj@swindj.UUCP (Alain Didierjean) writes: >In article <15983@cbmvax.commodore.com>, Ken Farinsky - CATS writes: >-> Have you considered sending bug reports (along with sample code) to SAS? >-> Or are you just going to complain that it is broken? Developers can >-> only fix problems that: 1) they know about, and 2) they can reproduce. >-> -- >Considering that the only practical and not too expensive way to get in >touch with SAS people from here is this very net, could we possibly get >an E-Mail address for SAS Institute Amiga C debugging crew... Or any >organisation that agree to be used as a relay of some sort... >This could possibly be useful to some american users too. >If not, I'm going to join the "SAS bug of the month club". Not very >efficient, but excellent for mental hygien.. I can only agree with Alain. Here's my bug of the month: Try compiling this: ----------------------------------- typedef int test; void test(int x) {} ----------------------------------- LC1 catches a GURU on this one. Anybody who doesn't have to call long distance overseas international terribly expensive is welcome to send a formal report on this. --- \_/ Mikael Karlsson, Lovsattersvagen 10, S-585 98 LINKOPING, SWEDEN V | micke@slaka.sirius.se | Absolut Software | micke@slaka.UUCP ~~~ | {mcvax,seismo}!sunic!liuida!slaka!micke