[PROJ] PROJ without proj.db

[Søren Holm]

Sorry for the noise Kristian .... turned out that there where multiple 
contexts and not all of them had the logger function set. Doh!!!

onsdag den 10. marts 2021 09.22.32 CET skrev Kristian Evers:
> Sorry, I completely misread that. Well, I guess my analysis is still valid
> although it doesn’t solve your problem :-)
> 
> I am not sure what’s going on then. Which version are you using?
> 
> /Kristian
> 
> > On 10 Mar 2021, at 09:18, Søren Holm <sgh at sgh.dk> wrote:
> > 
> > The function pointer is the third argument. I do indeed set the log
> > function.> 
> > onsdag den 10. marts 2021 09.14.34 CET skrev Kristian Evers:
> >> Here’s the source proj_log_func():
> >> 
> >> /************************************************************************
> >> *** **/ void proj_log_func (PJ_CONTEXT *ctx, void *app_data,
> >> PJ_LOG_FUNCTION logf) {
> >> /************************************************************************
> >> **
> >> **** Put a new logging function into P's context. The opaque object
> >> app_data is passed as first arg at each call to the logger
> >> *************************************************************************
> >> *** **/ if (nullptr==ctx)
> >> 
> >>        ctx = pj_get_default_ctx ();
> >>    
> >>    ctx->logger_app_data = app_data;
> >>    if (nullptr!=logf)
> >>    
> >>       ctx->logger = logf;
> >> 
> >> }
> >> 
> >> So, by putting nullptr as the log function you end up with the default
> >> logging function. Try setting up an actual logger function and pass a
> >> pointer to that.
> >> 
> >> I can see how you might think that a null pointer would achieve something
> >> else, though. The docs should clarify that or perhaps the functionality
> >> should be changed to let logf==nullptr explicitly mean “shut up”. The
> >> latter is likely the best move based on the fact that if you are happy
> >> with
> >> the default logger you probably wouldn’t call proj_log_func() in the
> >> first
> >> place.


-- 
Søren Holm