pmoore@cix.compulink.co.uk (Paul Moore) (05/18/91)
I have just downloaded patch 1 to GhostScript, and I am still having problems with it. I set up the !GS application, and copied in the new !RunImage. Then I tried to run the application. At first, I got nowhere, until I remembered someone mentioning that the wimpslot on the !Run file was too small (why wasn't this included in the patch?) Now, I am getting an address exception at &7FE4 (obviously wrong, as application space starts at &8000!) every time. What is going on here? My machine is an A310, with 4M of RAM and an 80M SCSI hard disk. My CLib is 3.66, but the same happens with 3.50. I have FPE 2.80 and Colours 0.52. I am getting very sick of problems with GhostScript. I have spent a lot of time and money (in download costs) trying to get this to work. Someone has previously mentioned that Alan, as moderator, should be testing things to make sure they work before posting them. I wholeheartedly agree. It should not be too difficult to set up an Archimedes in a "basic" configuration (No special modules loaded, CLib 3.50, other standard modules at "official" release levels, etc) and try to run an application. Nothing major, just does it run, and does it give reasonable results on any test files supplied (eg the PS directory in !GS) In addition, where sources are supplied, Alan should be testing that they compile cleanly. Here, I would recommend that ONLY the tools in the Acorn C (3.00) distribution be required. Ie, makes done by AMU, no use of LibFile and ObjLib (ESPECIALLY not the non-standard versions required in !GS, which support the -via keyword, and where ObjLib appears not to be needed). Also, I would appreciate it if people would include an empty O directory in the distribution. A minor point, but it irritates me when an initial make fails "Can't create O.xxxx"! By the way, I would normally expect a rebuild to be done by a simple "AMU", with no parameters, unless it is explicitly stated otherwise in a !ReadMe file, and I would prefer it if the makefile did not depend on OS variables, such as <C$RISC_OSLib> in !GS. While much of the above ought to be done by the original developer, in theory, I am aware that it is a real pain to do, in practice. (I know MY stuff never follows all of the above rules!) However, I'm afraid that I DO think it's Alan's responsibility to enforce these (or some similar) rules. That's what a moderator is for, after all. By the way, Alan, you should *publish* the rules under which you wish postings to work, to make your job easier! By the way, I feel that ensuring EXECUTABLES run first time is absolutely vital to the continued success of c.b.a, but I would not complain too much if I needed to do a bit of work to get compiles working (eg Makefile conversion, etc). But on the other hand, if I DO recompile, I expect the rebuilt version to work!!! Here, !GS again fails, as I get "No stack for trap handler" and similar messages... I would be interested in people's thoughts on the above, particularly any comments from Alan, on what HE sees the moderator's role as. I must admit, though, that unless there is a definite change in the level of reliability of c.b.a postings, I, personally, will probably stop bothering... Gustav. (pmoore@cix.compulink.co.uk)
gtoal@castle.ed.ac.uk (G Toal) (05/19/91)
In article <1991May18.154626.22804@demon.co.uk> Paul Moore <pmoore@cix.compulink.co.uk> writes:
...
And the galling thing is, Ghostscript 2.2 is about to be released from
the FSF in a few weeks! If David hasn't been disheartened by the whole
saga we'll have to do it all again! At least we know which mistakes to
avoid this time :-)
I pass on a message from our site admin: could you please stage large
posts by sending only 2 or 3 each day, rather than all at once. That's
how the grown-up groups do it, and you don't run the risk of pissing off
the world's site admins - the volume of news at the moment is causing
problems everywhere, which are exacerbated by the batchiness it shows
on certain days.
I concur with Gustav's suggestion for Alan to lay aside a kosher A4[124]0
with the public releases of Risc OS & C on it for testing; perhaps the
software QA department could offer him a few hours on their kit? I always
found it annoying in the past that developers in Acorn clearly didn't
use the public versions (even when I was one :-) )
Grahamdahe@cl.cam.ac.uk (David Elworthy) (05/20/91)
In article <1991May18.154626.22804@demon.co.uk> Paul Moore <pmoore@cix.compulink.co.uk> writes: > until I remembered someone mentioning that the wimpslot on the !Run file > was too small (why wasn't this included in the patch?) It was certainly in the patch I sent to Alan - a change to 220k. > Now, I am getting >an address exception at &7FE4 (obviously wrong, as application space >starts at &8000!) every time. > >What is going on here? My machine is an A310, with 4M of RAM and an 80M >SCSI hard disk. My CLib is 3.66, but the same happens with 3.50. I have >FPE 2.80 and Colours 0.52. No idea what is causing this - I tested the release on a vanilla A310 with one floppy and CLib 3.66, and it worked. That is really all I can do. > By the way, I would normally >expect a rebuild to be done by a simple "AMU", with no parameters, unless >it is explicitly stated otherwise in a !ReadMe file, and I would prefer >it if the makefile did not depend on OS variables, such as <C$RISC_OSLib> >in !GS. The development was not done with, but I take your point about mentioning this in the release. Concerning the use of C$RISC_OSLib, you either rely on system variables, which are defined by typing *set once, or you rely on the disc structure of the machine it is being compiled on, which potentially needs to be changed all over the place. >While much of the above ought to be done by the original developer, in >theory, I am aware that it is a real pain to do, in practice. (I know MY >stuff never follows all of the above rules!) Then cut the hypocrisy. You are getting essentially for free something which took similar effort and skills to what I used to be paid a good salary for. If you don't like the result, then you have the sources - go ahead and produce your own version. >Gustav. (pmoore@cix.compulink.co.uk) -- david