Simon Slavin wrote and quoted:
>> Would the output of 'PRAGMA compile_options;' be sufficient ?
>
> That's a cute and useful feature I had forgotten. It would be sufficient to
show the preprocessor variable values that were in effect as the amalgamation was
compiled. However, for something that looks like either a SQLite code error or a
gcc code generation error, the optimization settings are also likely to be needed,
and those are not susceptible to that PRAGMA's behavior.
Don't forget this isn't is own weird project which may be using the API in an
unusual way. It's the shell tool. Possibly even a the version downloaded from
the SQLite web site. Given how clear the bug is, with the addition of the
bug-reporter's 'PRAGMA compile_options;', I think that's enough information to
let the team try to reproduce the behaviour and see what happens.
By the way, this shell tool included with a recent version of OS X does not
have the problem.
I don't want to get into a big debate over this, (in part because it's
an indeterminate sort of issue), but I am responding so that the O.P. is
not induced to ignore what I still think was good advice.
Of course "the team" can try to reproduce, something that can try with
whatever repro info they have. And, as I also indicated, the
preprocessor settings may be enough. If they are not, if the problem in
fact arises with some strange combination of the seemingly countless
optimization and other code generation options gcc has, the O.P. will
ultimately have to discover that himself or provide those options to
"the team". It would be more efficient for whoever does this little
investigation to have full repro data. If they can only guess at the
compilation flags, they do not have reproducibility due to the
practically unbounded combinations, and that investigation will have to
be broken into more sessions while the data I suggest be provided is
obtained. Getting that data is not some big effort. If the O.P. is an
experienced developer, he will know just how to get it. If not, he can
either state he is using defaults or reflect upon whether playing with
intriguing gcc flags is the best course.
To the O.P.: As someone who has seen a few code generation problems, I
can state without doubt that optimization or other code generation
settings may very well be relevant here, and that providing them up
front will help resolve this if that is the case.
To Simon: I do not claim here that these settings *are* the problem,
just that if they are, they will need to be made evident to enable
progress to be made toward a solution.
--
Larry Brasfield
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users