[uClibc]uClibc features?

[David Schleef]

On Sat, Jul 21, 2001 at 02:01:47AM -0700, David Schleef wrote:
> 
> However, the process of thread-safing a library has at least one
> advantage that is interesting in uClibc's case -- basically every
> global variable can be moved from the data section to the static
> data section, causing processes to share more RAM, and potentially
> saving a small amount of RAM (4-20 kB as a guess) per process.  This
> would be a noble undertaking in itself.

I checked up on this, and got very excited when I found out the
.data section for libuClibc.so on powerpc is about 12kB.  Then I
got confused when I found that it's about 1.5kB on i386.  It seems
like there's a bug in gcc-2.95.4 on powerpc that sometimes puts
constant character strings in .data and sometimes in .rodata/.sdata,
depending on the number of arguments of the function that is called.
(Not "const char *" vs. "char *", or something you might expect.)

Does anyone have experience with bugs like this?  I'll look into
it more and post some test code sometime when it's not a sunny
Saturday afternoon.



dave...