Re: [Chicken-users] Re: sqlite3 issues

[elf]

the thing that got it to work (the others are mostly incidental except for
initialize) was changing the require-extension format used in the file.

basically, i was shotgunning it, because it would kill the interpreter
immediately upon trying to load it, and even valgrind with all the fruit wasnt
helpful in figuring out what was going on.  it turned out to be from what
i consider a problem with the compiler, namely that
require-extensions/use/uses arent verified at compile time, and if the
extensions do not in fact exist (or cant be loaded, or whatnot), then it
crashes the interpreter.

there also seems to be some degree of confusion involving the (unit)
declaration, the register-feature! functions, and the supported types
for use vs uses vs require-extension, which dont all match up, for some reason.

-elf


On Wed, 8 Nov 2006, Thomas Chust wrote:

> On Wed, 8 Nov 2006, elf wrote:
>
> > heres the files with the modifications i made to get it to work right.  i
> > also removed the need to install the scm file cause there are no syntax
> > macros anymore, and got rid of the initialize fucntion which always yelped
> > because it was overriding things, and various other shtuff.  hope this 
> > helps.
>
> Hello,
>
> well, it was completely unnecessary to install the .scm file in the egg
> repository in the first place, this wouldn't have needed any
> modifications. It's my fault that the install command included the source
> file, but the file would just lie around in the egg repository and do
> nothing, because the extension was never marked as syntax.
>
> Removing my own enumeration support macros and replacing them with the
> builtin ones may be a good idea -- my own macros come from a time when the
> CHICKEN builtin ones behaved completely differently and I didn't like
> their behaviour. But then again, as that change does not improve
> performance I tend to not change a running system.
>
> Finally, removing the initialize function or not exporting it is probably
> a bad idea. It is strictly necessary to get working pointers into the
> wrapper objects and all hell in the form of random segfaults and the like
> may break loose if you remove it. If the code still works without this
> function, that's a pure coincidence. The warning emitted by the compiler
> seems to be a necessary evil ;-)
>
> Considering everything, I can't see that you have made any significant
> change in the source or compiler flags. If it works after the changes I
> don't see a reason why it shouldn't have worked before. The egg also still
> runs fine on my machine and a test box running Linux, both with an updated
> SQLite3 installation and unmodified egg source.
>
> Sorry, but I'm totally clueless as to what is causing your problems :-(
>
> cu,
> Thomas
>
>

-----
Humans always do the most intelligent thing after every stupid alternative
has failed.                   -- R. Buckminster Fuller

A year spent in artificial intelligence is enough to make one believe in G-d.
                              -- Alan Perlis