>>
Can you compile and run a "Hello World" example in C or C++ yet?
>
To build OpenWatcom, where do you get the OpenWatcom binaries?
>
Is anybody considering to "port" OpenWatcom to Linux?
>
I noticed that the Windows and OS/2 compilers will be included. What about DOS and the DOS 32-bit Extender?
>
Are the compilers for other platforms going to be included?
Yes and currently are.
Not ANSI conformant and export keyword not implemented.
The only C++ compiler that is fully conformant and the first that supports export is Edison Design Group. See comp.lang.c++.moderated.
Those are kind words but they are not accurate:-) I could list areas where the compiler could improve, no problem; areas where I know of bugs or where code quality could improve. Note I am a Watcom advocate but I like to call a spade a spade and measure truthflly against opponents. And I do think MS uses VC++ for Win2K.
I used Watcom C 7.0 in 1987 and had swapped to it because Microsoft C 5.1 at the time was poor. Over the years, Watcom improved. But over the years Microsoft improved and its current C & C++ compiler, version 6.0, IMO produces code that more often than not is better than Watcom particularly in the area of floating-point. With 11.0c, an important floating-point switch bug was fixed in the C++ compiler so I would like to see Watcom 11.0c C/C++ retested against benchmarks that are quoted here.
Version 7.0 of VC++ is interesting as well. It is much more compliant to the ISO C++ 99 standard than version 6.0. Truly MS have done a lot of work.
So, IMO, there is lot of work for the OpenWatcom team (I am a member) to make the compiler faster and as compliant (if not more so - I am thinking of member templates which VC++7.0 will _NOT_ implement) than VC++.
I still don't know where this "donation" business comes from. As far as I am concerned it is fiction. And as I have been involved with the Open Sourcing of Watcom from start to finish, before even SciTech got involved, I know.
And we don't to prove anything to anyone. I am just glad that existing customers, some of whom I have known 6 years, have a better product in their hands than 11.0b.
>> But because these folks aren't a big corporation it's OK for them to scam people out of donations.
What donations, what scam? First I know of it.
>> So Sybase.... further on down the road
Except you are flatly wrong. Firstly, the fact is Sybase has plenty of people internal to maintain the compiler and tools, they need never open source it at all, period. Effectively the compiler would become internal to Sybase. They could still do that, why have any potential hassle from open sourcing it? I know this, because most of the fixes in 11.0c are a direct result of emailed conversations between myself and Watcom R&D in Waterloo, Canada since Feb 1998.
Secondly, all the lawyer conversations have been ironed out as of 2 months ago.
Thirdly, your hypothesis is wrong. Unlike you, I have talked face-to-face with Sybase director responsible with open sourcing the compiler. His interest was that the existing customer base would have a means of supporting their applications and investment in Watcom C/C++ & Fortran. Now it maybe that Sybase does profit from work on the compiler but hey, that is their right. It would not surprise me if the license is FreeBSD rather than GPL. Everyone benefits.
Well it is somewhat better. Hello World is in size
16-bit DOS: 61444
32-bit DOS: 92504
Win32: 81920
Looking at the map files, we can improve the granularity of OBJ files. So there is room for improvement.
>>Until source has been released I consider the project vaporware.
Well that is your opinion, but as I am on the OpenWatcom team and also in contact with Watcom R&D, Waterloo, Canada, your opinion is wrong. I have seen things move forward internally.
>>But why do they......until they can sort it out
That is exactly what is happening:-). 11.0c consists of files that are only 100% Sybase owned + any agreements we make with various 3rd party authors.
You have some excellent points. But commenting on a few things:
>>>
The C-parser was buggy.
>>>
Not my general experience at all. And I have used every version from 7.0 to 11.0. But some versions were better than others. For example the initial version 10.0 was flaky as was 11.0. But versions 10.6 and 11.0b are very solid.
>>
The compiler was buggy.
>>
Hand on heart, I can say that that is not my experience with most versions. I have seen compiler bugs with 11.0, 11.0a and 11.0b and know what there are because of what customers reported.
>>
Inserting assembly code used a totally different syntax than Borland C.
From version 11.0, __asm blocks can be used.
Best ask those who are still using 32-bit DOS Extenders. There are quite a few that ask for help in the Watcom newsgroups. I don't know why they are using them
Out-of-the-box with pre-Beta 11.0c and no previous version of Watcom, you can produce 16-bit DOS targets. That is all.
With the final 11.0c patch you should be able build 16 and 32-bit DOS targets (is likely to include DOS4GW.EXE)
For other targets, you need 3rd-party files and libraries but from what I have read, OS/2 and Win32 executables are not problem with the latest Development Kit from IBM and Platfrom SDK from Microsoft.
You should not have a problem with Watcom Fortran
>>
Yet they still haven't released a single line of source. They haven't even decided on a license. All they did is release a binary patch.
>
Why don't they just rip out the stuff they can't distribute, and publish the incomplete source?
We don't have to prove a damn thing. As it stands, the tools are based on 10.6, a previous version of the compiler. That will _never_ be open sourced. But 11.0c will be.
11.0c is for existing customers, OpenWatcom 1.0 will for everyone else.
>>Perhaps this time next year they'll release >>another beta of this patch
Won't happen
>>get a little more free publicity from Slashdot
I don't care if we never get any publicity from Slashdot. AFAIK, we never sought it.
>> and get a few more dollars in donations.
What is all this talk of dollars? News to me.
>> How exactly does this project qualify as open >> source?
It doesn't. There is a schedule. Ever since 11.0b was released in Feb 1998, there have been a number of fixes by Watcom R&D in Waterloo. I have made a tiny few myself. 11.0c is for _existing customers_ and it includes all the fixes to date that have occured since 11.0b.
After this comes OpenWatcom 1.0. There are number of things that need doing (cleaning up the source a bit which we have a little). At this point you will be able to see the source.
>> I wonder if this will be released under a GPL license?
Possible but it is unlikely Sybase will do that.
>> Or whether the two compilers themselves will >> ultimately be merged into one product?
Very very unlikely. For example, Watcom is has passed the first few arguments for functions in registers and that is unlikely to change. To do so would alienate existing customers and would invove a complete rewrite of the compiler.
I can see that there will probably a fruitful exchange of ideas but not code.
Slashdot Mirror
>> Can you compile and run a "Hello World" example in C or C++ yet? > To build OpenWatcom, where do you get the OpenWatcom binaries? > Is anybody considering to "port" OpenWatcom to Linux? > I noticed that the Windows and OS/2 compilers will be included. What about DOS and the DOS 32-bit Extender? > Are the compilers for other platforms going to be included? Yes and currently are.
Not ANSI conformant and export keyword not implemented. The only C++ compiler that is fully conformant and the first that supports export is Edison Design Group. See comp.lang.c++.moderated.
Those are kind words but they are not accurate :-) I could list areas where the compiler could improve, no problem; areas where I know of bugs or where code quality could improve. Note I am a Watcom advocate but I like to call a spade a spade and measure truthflly against opponents. And I do think MS uses VC++ for Win2K.
I used Watcom C 7.0 in 1987 and had swapped to it because Microsoft C 5.1 at the time was poor. Over the years, Watcom improved. But over the years Microsoft improved and its current C & C++ compiler, version 6.0, IMO produces code that more often than not is better than Watcom particularly in the area of floating-point. With 11.0c, an important floating-point switch bug was fixed in the C++ compiler so I would like to see Watcom 11.0c C/C++ retested against benchmarks that are quoted here.
Version 7.0 of VC++ is interesting as well. It is much more compliant to the ISO C++ 99 standard than version 6.0. Truly MS have done a lot of work.
So, IMO, there is lot of work for the OpenWatcom team (I am a member) to make the compiler faster and as compliant (if not more so - I am thinking of member templates which VC++7.0 will _NOT_ implement) than VC++.
>>I've been following the posts here, and I don't know where the "donations" bit came up from either
:-)
Scitech emailed me. If you look at
http://www.openwatcom.org/download.html
there it is. Embarassing not to know (when I usually know beforehand)
Read the statement carefully. The key word is "distribution".
I still don't know where this "donation" business comes from. As far as I am concerned it is fiction. And as I have been involved with the Open Sourcing of Watcom from start to finish, before even SciTech got involved, I know.
:-)
And we don't to prove anything to anyone. I am just glad that existing customers, some of whom I have known 6 years, have a better product in their hands than 11.0b.
But hey, good times are ahead
>> Still, the only thing they've done so far is make ...
No we haven't. If we did so, we would make 11.0b available. As it is, 11.0c is made available which contains 3 years worth of fixes.
Have a look at the file
ftp://ftp.openwatcom.org/watcom/c_readme.txt
>> But because these folks aren't a big corporation it's OK for them to scam people out of donations.
.... further on down the road
What donations, what scam? First I know of it.
>> So Sybase
Except you are flatly wrong. Firstly, the fact is Sybase has plenty of people internal to maintain the compiler and tools, they need never open source it at all, period. Effectively the compiler would become internal to Sybase. They could still do that, why have any potential hassle from open sourcing it? I know this, because most of the fixes in 11.0c are a direct result of emailed conversations between myself and Watcom R&D in Waterloo, Canada since Feb 1998.
Secondly, all the lawyer conversations have been ironed out as of 2 months ago.
Thirdly, your hypothesis is wrong. Unlike you, I have talked face-to-face with Sybase director responsible with open sourcing the compiler. His interest was that the existing customer base would have a means of supporting their applications and investment in Watcom C/C++ & Fortran. Now it maybe that Sybase does profit from work on the compiler but hey, that is their right. It would not surprise me if the license is FreeBSD rather than GPL. Everyone benefits.
Hope this clarifies matters
Cordially
Well it is somewhat better. Hello World is in size 16-bit DOS: 61444 32-bit DOS: 92504 Win32: 81920 Looking at the map files, we can improve the granularity of OBJ files. So there is room for improvement.
>>Who says they ever will be solved?
...until they can sort it out
:-). 11.0c consists of files that are only 100% Sybase owned + any agreements we make with various 3rd party authors.
They have been solved
>>Until source has been released I consider the project vaporware.
Well that is your opinion, but as I am on the OpenWatcom team and also in contact with Watcom R&D, Waterloo, Canada, your opinion is wrong. I have seen things move forward internally.
>>But why do they...
That is exactly what is happening
Fair enough, I understand what you say. When we release OpenWatcom 1.0 you can try a copy then.
I had no choice because this infernal SlashDot would not allow me my own name
You have some excellent points. But commenting on a few things: >>> The C-parser was buggy. >>> Not my general experience at all. And I have used every version from 7.0 to 11.0. But some versions were better than others. For example the initial version 10.0 was flaky as was 11.0. But versions 10.6 and 11.0b are very solid. >> The compiler was buggy. >> Hand on heart, I can say that that is not my experience with most versions. I have seen compiler bugs with 11.0, 11.0a and 11.0b and know what there are because of what customers reported. >> Inserting assembly code used a totally different syntax than Borland C. From version 11.0, __asm blocks can be used.
Best ask those who are still using 32-bit DOS Extenders. There are quite a few that ask for help in the Watcom newsgroups. I don't know why they are using them
Out-of-the-box with pre-Beta 11.0c and no previous version of Watcom, you can produce 16-bit DOS targets. That is all. With the final 11.0c patch you should be able build 16 and 32-bit DOS targets (is likely to include DOS4GW.EXE) For other targets, you need 3rd-party files and libraries but from what I have read, OS/2 and Win32 executables are not problem with the latest Development Kit from IBM and Platfrom SDK from Microsoft. You should not have a problem with Watcom Fortran
>> Yet they still haven't released a single line of source. They haven't even decided on a license. All they did is release a binary patch. > Why don't they just rip out the stuff they can't distribute, and publish the incomplete source? We don't have to prove a damn thing. As it stands, the tools are based on 10.6, a previous version of the compiler. That will _never_ be open sourced. But 11.0c will be. 11.0c is for existing customers, OpenWatcom 1.0 will for everyone else.
>>Perhaps this time next year they'll release >>another beta of this patch Won't happen >>get a little more free publicity from Slashdot I don't care if we never get any publicity from Slashdot. AFAIK, we never sought it. >> and get a few more dollars in donations. What is all this talk of dollars? News to me. >> How exactly does this project qualify as open >> source? It doesn't. There is a schedule. Ever since 11.0b was released in Feb 1998, there have been a number of fixes by Watcom R&D in Waterloo. I have made a tiny few myself. 11.0c is for _existing customers_ and it includes all the fixes to date that have occured since 11.0b. After this comes OpenWatcom 1.0. There are number of things that need doing (cleaning up the source a bit which we have a little). At this point you will be able to see the source.
>> I wonder if this will be released under a GPL license? Possible but it is unlikely Sybase will do that. >> Or whether the two compilers themselves will >> ultimately be merged into one product? Very very unlikely. For example, Watcom is has passed the first few arguments for functions in registers and that is unlikely to change. To do so would alienate existing customers and would invove a complete rewrite of the compiler. I can see that there will probably a fruitful exchange of ideas but not code.
It is free. Since I am on the Openwatcom development team, I know what I am talking about