The less patch is in.

[Rob Landley]

On Saturday 17 September 2005 16:06, Rob Sullivan wrote:

You asked about size reduction guidelines.  First of all, "make sizes" is your 
friend.

> - Replaced the free_flines(), data_readlines(), etc sequence with one
> function: reinitalise().
> *** no effect on size ***

Not having seen what the actual patch was for this change by itself, I have no 
idea...

> - Fixed up flag detection to be more Busybox-like.
> *** saves 94 bytes ***
>
> - Changed buffer[100][256] to **buffer and also changed all array
> assignments involving the buffer to
> pointer assignments.
> *** saves 1824 bytes ***

How are you getting these numbers?  Running make sizes shows you the actual 
sizes of the components.  Looking at the .o files or the final binary is 
misleading.

> - Changed *flines[MAXLINES] into **flines - this reduces memory usage, but
> does increase the size of
> the less binary somewhat.
> *** adds 2026 bytes ***

Again, not having seen what the actual patch was without the changes all mixed 
together, I have no idea what you did.

> The last two changes make me feel a bit nervous - why would changing from
> an array to pointers save or add so much respectively? Perhaps something's
> gone wrong with my flines change...

Most likely the extra size was due to an extra function to be linked in from 
libbb, which would already be there with more applets in the binary anyway.  
I don't know what you did...

I'm going to check this in before thumping on it myself, but it looks like 
you've got a memory leak if anybody calls full_repaint() (which calls 
data_readlines() which doesn't free flines[] because reinitialize() is doing 
that, but this codepath doesn't go through that...)

Rob