Rockbox

I do not use the databse but I think it's a very nice feature and should be committed.

merge my ui patch, not really tested and diff has extra crud…. laptop battery going to run out before i can finish today

   changes.diff (46.1 KiB)

better patch, my end is done, looks like your tagcache.c changes need a bit of cleaning up.

Last thing thats needed on the ui side is to fix skin lists to do indenting (not trivial so too hard for my current extremely overtired state)

   changes.diff (23.1 KiB)

Thanks for porting the UI!

Which of my tagcache.c changes are you refering to? You seem to have added some which need cleaning up (the if (ignore) closedir()…, and #if 0 stuff).

A few other remarks:
a) IMO the search roots shouldn't be a separate file. It's a setting after all.
b) Shouldn't your new fancy UI be more generic and not a tagcache specific file (going by the file name)?
c) fast_readline looks wrong buggy. The callback is not called for every line, but for the first one of each read (the size of which depends on the buffer passed to it).

I haven't actually tested yet, just looked at the diff. Will try out later

a) if you're talking about tagcache_config.c then no, it should stay seperate, 1) it makes it easier to becomee generic later, and 2) its 400 lines which isnt relelvant to anything else
b) deal with that later if anything else wants to use it
c) ?

ah, maybe i didnt merge our ptches cleanly, I'll go through it again if you say those changes arent yours.

a) I meant the TAGCACHE_SEARCHROOTS file where the dirs are saved to with.
b) should be easy enough to do now. The tagcache specific part can move to tagcache.c. Isn't that only one function anyway, the rest being completely generic? My fear is that nothing is going to want to use it because it's quite hidden behind a file name that suggests being quite tagcache-specific (devs just don't notice).
c): The fast_readline() function looks bogus to me, as if the callback isn't actually called for each line of the file. It's one call to the callback per read(). One read() can get multiple lines at once, though.

a) ah, well that gives more flexibility than using global_settings (and wastes less ram doing so), the disk needs to be spinning when the db is built anyway so no penalty there.
b) meh, nothing will use it because nothing else needs to really. if someone wants to split it later they can
c) I honestly have no idea what you're saying, if fast_readling() is broken its not new in this patch.

a) I don't see how it gives more flexibility. But I guess your code is already done so it assumes one item per line as you often use readline().
b) I'm sure autoresume wants it
c) yes, not new in this patch. It just looks wrong to me unrelated of this patch.

fast_readline is correct.. the read() call will continue to fail once it hits the end of the file but the loop wont exit until if ( (p = strchr(buf, '\n')) != NULL) fails also.

Here's an update. It takes jdgordon's work, but removes the extra configuration file and shows the root folder ("/") so that the entire filesystem is selectable.

   0001-Configurable-database-sc... (28.7 KiB)