Re: [Mingw-users] DLLs and Exceptions

[Oscar F. <of...@wa...>]

Danny Smith <dan...@ya...> writes:

[snip]

> What do users want.  Revert to sjlj, which works, but incurs a
> performance penalty comapared to Dwarf2?

Just in case somebody doesn't know: sjlj exceptions are used on MinGW
gcc versions up to 3.2. In case we revert to sjlj, we go back wrt
MinGW gcc 3.2, not gcc 2.x. It would be useful to know how much is the
performance penalty on our apps, although I wouldn't give away
stability for obtaining performance on any case.

> Or stick with Dwarf2 and hope that the proverbial soemone will fix
> the known bugs in the near future.

We discussed that on the past, but here I go again: 

Some people wants DWARF2 because it's "cool" (and it is, although many
people will never know it is there), others want performance at any
cost. Finally, a group of users _requires_ stability 'cause they want
to use MinGW for production code. For this group, asking "Do you want
the coolest technology or something that works?" makes no sense. I
belong to the last group, but I respect the others.

If it is possible to build gcc from sources with the desired option,
the problem vanishes as far as you are motivated enough to do it. (I am)

But then, the real questions are: For those that doesn't want to hear
about "compiling the compiler", should MinGW focus on stability or on
cutting-edge technology, even when it is known that doesn't work?
Shoud we wait until a feature really works before introducing it (as
GCC/Binutils does) or should we promote MinGW as an experimental
project? Is MinGW mainly a tool for creating software of something
geared towards people interested on compiler construction?

I know the answer to the above is "both", but I'm afraid MinGW doesn't
have enough resources to be both things, so we need to move the
balance to one side or the other. Let's see what people says.

FWIW, my vote is MinGW == tool == stability

-- 
Oscar