Richard Stallman is responsible for the shrinking economy

... snip ...

Because a GPLd version is a dynamic thing. It can, and will, be improved with time. The commercial ones won't get altered unless someone at that commercial entity is convinced of the need. It does mean that the maintainers have to be careful not to accept silly complaints.

... snip ...

Fine. It can still be used to form an initial GPL test suite for standard C. The first problem is to delete things that are non-standard, and are not tracable to the standard. Then the development can be left to the Open Source world.

Testing to the standard ONLY would be an incredibly helpful thing.

All test suites are "small" compared to the space of valid programs.

The kind of bugs which most often bite real-world code tend to be interactions between highly specific optimisations which you could never expect to catch with a test suite.

A specific example: one version of MSVC with optimisations enabled screwed up a loop to initialise a 4-element array. If the number of elements in the array was less than 4 or more than 4 it worked fine.

Comparing the assembler output for various cases showed that the problem was that it unrolled the loop if it had 4 elements or less, and left it as a loop for more than 4 elements. The 4-element case was a hybrid: most of the loop was repeated 4 times, but one specific part only appeared in the first copy.

The only test-suite which will trap a case like this is one which includes specific tests for previously-discovered bugs.

So who will do the work?

GO on then. It is a good idea.

Not sure about that. Given the size of a formal test suite, some 80,000 tests, and the reputation or rather lack of it amongst those who understand test suites of the current GCC "test" suite youare better off starting again.

Besides all the tests would beet to be written in a standard format so you would need to re-write the current tests anyway,.

That is what Plum-hall and Perennial do now. They do language compliance. You will need additional tests. You could start with Paranoia.

I think you will find you need C95 and C99, by the time it is completed you will also need C1*

In message , Rainer Buchty writes

No I don't know that. Some one said what about X and Y and promptly got told both were available.

I have had requests for obsolete software before and normally I have found you can still get your hands on it one way or another.

Again I have been told that there is a problem with dongles yet I have had compiler companies fix problems with dongles on compilers up to 18 years old. In that case it was a Franklin compiler and it was sorted out 2 months ago for Anton who posts on this NG.

I have had IAR sort out an obsolete and smashed dongle on a compiler that was about 8 years old. I don't know of a case where the problem has not been solved.

I agree. However I suspect that you would be better off starting from scratch than trying to use the current test suite

The current industry standard test suites are about 80,000 tests

Could you kindly send some references on that decision. It would be woth a look ...

- Regards, Sachin

Commercial compilers are tested against Plum-Hall or perennial for language compliance. They are then further tested with several other test suites for the mid and back ends including regression tests. This testing covers a hell of a lot in a structured and repeatable manner. Obviously it does not cover every eventuality.

The GCC has a partial and non structured regression test and nothing else.

So your preference is for something with a 5% ad-hoc test coverage over something with 80% structured test coverage?

I agree with Chuck. Perhaps we could play "All the President's Men."

"If you do not answer before next Monday, it´s Cosmic C or Bytecraft. If you don't answer before Tuesday, it's Cosmic C."

:)

"Ignacio G.T." wrote in news:49e88896$0$29138$ snipped-for-privacy@read.cnntp.org:

I still have to deal with the company, so I'm going to continue being cautious, however I will say it was not Cosmic C or Bytecraft. My experience with both of them has been excellent.

I don't have any objection to paying for software and support, as long as I am able to get acceptable responses. I use open source tools but usually get commercial support for them (Cygnus, when they were there, Microcross more recently).

However, I am very uncomfortable with tools that are "locked" in any way, so far I have a collection of over a dozen tools that I can no longer use for various reasons (the license manager won't run, the dongle needs a slow parallel port, etc.) and while the suppliers have usually been prepared to supply a copy of a previous version of their tools, they have almost always done it "as a favor" with no support, and no guarantees, and often, no success, while I have always been able to find someone with the knowledge to update an open source tool. I recently had to get an old version of gdb (using code from GCC 2.95) to connect to a remote system on ARCNET, from a computer running Windows, official Vista ARCNET drivers are rare, and Vista has some interesting security issues.

The Cygnus folks now make up part of the GES (Global Engineering Services) group at Red Hat, and you can still buy support from us.

Of course. I don't look at assembler code when there is no problem. In this case the function returned wrong results. The compiler maker confirmed that this is a bug, and we had the fix a few days later. No need to speculate what I could have done wrong in my analysis here.

Stefan

Sorry, I don't accept either source.

Okay, you can get Turbo C from Borland's "Museum" server. Actually, that's where I got it from. You can get one specific version of it. And, as far as I know, you get it under a license forbidding commercial use. Nothing I would base my important projects on. Other old Borland products cannot be obtained at all, for example Borland Pascal 7.

And if Walter knows someone who knows where to get a copy of Coherent with questionable legal status, that doesn't count as available software as well. Probably someone of us also knows someone whose grandpa wrote the software that flew Apollo 11 to the moon. Even if he still has the listings in his cabinet, I wouldn't say Apollo 11 sources are "available".

Stefan

IANAL, I just follow news with half an eye... A little searching turned up this section of the German Code of Commerce: (and the following)

They have to pay you _half_ your salary for the duration of the "no competition" contract. Still a good work/cash ratio :-)

Stefan

Don't you think releasing that information would also be violating their license?

Like many people, including me, it seems that Stan is happy to publicly name companies that have impressed him, but reluctant to put companies reputations at risk by bad publicity. If I have had a bad experience with a company, I would be careful in my own dealings with them, and may mention it to customers if the topic comes up. But I'm not going to tell the world about it - it could easily be an isolated incident, and I would not want anyone to suffer unnecessarily. Just like people, a company is innocent until *proven* guilty - a single bad incident is not enough.

Sorry, I took it for granted that people would be familiar with QT - especially people interested in open source development. But in hindsight that assumption was not reasonable (perhaps I know more about it since it's from a Norwegian company).

QT is a C++ class library for cross-platform development, especially for gui development. The developers also supply a range of tools that go with the library. The library has for many years been available in a GPL version for Linux (this is the foundation for KDE), and closed-licence paid-for version for Windows and Linux (and possibly other systems - I don't know the details). There are also several embedded versions aimed at fancy telephones, etc.

Of course I prefer well-tested code. I just don't subscribe to your view that gcc is "not tested at all", and that every single small change breaks it completely to need a full re-qualification. gcc compiles tons of practically-used code each day, so chances are good it groks my practically-used code as well (and because I'm a hobbyist C++ language lawyer, I stress it with template-metaprogramming stuff as well, which current versions also survive). Thus, gcc is mostly tested by the "given enough eyeballs, all bugs are shallow" principle.

Stefan

The laws in the US on the subject of "non-compete" clauses seems to vary a bit across states. In my state, they simply have no weight at all even if you sign them. I'm told I can sign the contract today, quit the job tomorrow, and start working on a competing product the next day. They can squirm, yell at me about the clause, plead, beg, and otherwise wish it weren't so, but these clauses have no necessary effect on me, in and of themselves.

(Signed documents from technical employees can make potential arms-length financers happy, though, because not all such investors are aware of all laws in all states and their own states may treat this differently so they may assume things incorrectly in my case.)

Sometimes, this is called "right to work" in law.

I'm speaking from personal experience where I hired a lawyer for exactly this issue. I believe I'm interpreting the advice (and results), correctly. It appears to be a reality in this state.

Jon

This is why German employers have to pay you a compensation. Thus, you still have a contract, which simply doesn't require a return in the form of appearing in the office at regular intervals.

The Code of Commerce outlaws contracts that say "after the termination of this contract, you still are not allowed to work for the competition" without compensation. Other post-termination clauses do not require compensation. For example, engineers are usually forbidden to talk about previous employers' trade secrets, but don't get any money in return for being silent.

To ensure this right to work, the employer may not restrict you too much, otherwise the clause is void (this is in §74a following the one I quoted). That is, he may not disallow me developing any embedded software, or entering any company within Saxony.

Stefan