----- Original Message ----- From: "Frank Heckenbach" ih8mj@fjf.gnu.de To: gpc@gnu.de Sent: Wednesday, April 07, 2004 10:52 AM Subject: Re: GPC ISO 7185 compliance data
Scott Moore wrote:
Made changes listed above.
I can't interest you in supplying any comments for the "originator comments" field in the "Verified compilers" section ? It was intended that you would have the opportunity to rebut anything I said, explain problems, or add other comments.
Since what you wrote is basically all positive, I don't have much opportunity for rebuttal. :-)
I could point out that the MaxInt value and default field widths can depend on the processor type and dialect options, but since you stated the processor type and it's clear from the context that it's referring to ISO 7185 mode, it may not be relevant to mention this.
: 2. The ISO 7185 Version of Pascal-s compiled sucessfully and ran a : small program. However,. it failed to run the test suite for : Pascal-s. I plan to determine why this occurred as time permits.
Of course, I can't tell what's going wrong here. It may or may not be a GPC bug ...
Pascal-s does run, but the test suite I give it is pretty complete, and apparently has found a corner case. It runs on other ISO 7185 Pascal implementations. Its probally not fair that I mention it, but right now I just don't have the time to track it down. However, I will get to it and let you all know what is going on.
: [...] heard from other sources that have familiarity with the BSO : test that it has several gaps in ISO 7185 coverage.
BTW, how completely do your tests cover ISO 7185? (This is not meant to question the validity of the tests, quite the contrary. How sure can we be of ISO 7185 compliance if GPC will pass your tests?)
Frank
There's probally no real good answer to that question. The test is about 13 years old now, and incorporates all the ISO questions that have come up (in comp.lang.pascal.ansi-iso, for example). It is a "positive" test, meaning that it performs no checking on if invalid programs or program structures are flagged, and it performs little or no "stress" type tests, like checking how many nested structures can be handled, etc.
In my own experience, with IP Pascal, the test was run, and then several more bugs were found after having sucessfully run the ISO 7185 test and proceeding to compile various large programs (like the compiler itself). These bugs were divided into bugs due to IP Pascal extentions to the standard, and "corner" cases, such as performing an operation in such a way as to trip an optimization that was not correct. I have never seen an ISO 7185 construct go unchecked by the test suite, but thats by definition. If I were to find such a thing, it would go into the test suite. So far, however, the test suite has always been ahead of the programs I write in terms of the ISO features used.
The BSO tests are supposed to be very good at negative testing, ie, they apparently evaluate how good the compiler is at rejecting bad constructs. My ISO tests are very weak at this.
I have been trying, for years in fact, to get an agreement going with Tony Heathrington of Prospero software. They aquired the rights to the BSO test suite, and Tony has been talking for years about giving better price terms now that the BSO test is no longer "mainstream". I would personally like IP Pascal to pass the BSO tests as well, and I think that would be a good goal for GPC. I think you might be able to get that for free considering the nature of GPC, with proper garantees that you won't distribute it.
However, Prospero is not answering their email, and I fear they may be going down for the count.
-- Frank Heckenbach, frank@g-n-u.de, http://fjf.gnu.de/, 7977168E GPC To-Do list, latest features, fixed bugs: http://www.gnu-pascal.de/todo.html GPC download signing key: 51FF C1F0 1A77 C6C2 4482 4DDC 117A 9773 7F88 1707