[flashrom] [RFC] libflashrom design choices

[Carl-Daniel Hailfinger]

Hi,

since the request to turn flashrom into a library pops up from time to
time, we have to make sure the flashrom code doesn't have any nasty
surprises.

- We have to decide on a design for fatal errors after initalization is
over. Some code uses exit(1) for any fatal errors after init, but that
it not really a valid choice for a library. After all, the caller may
want to clean up some stuff outside libflashrom.
- Kill all uses of exit(1) inside init code. The programmer_init code
has proper error handling and the caller must check its return code. I
have a patch for this.
- Memory leaks. We didn't have to care because everything was freed
automatically upon exit. I have a patch for the leaks I found from a
cursory glance at the code.
- Global error variable?
- Grouping all global variables into a single struct to avoid namespace
pollution.
- Adding struct flashchip to every function or relying on a global variable.
- Thread safety?
- Which symbols (variables/functions) should be exported in the library
interface?
- Consistent prefix for exported functions (namespace pollution)?
- message level functions (print_debug/print_warn/print_err/print)
instead of printf and friends so the library user can decide how/if to
display these messages.

After all the stuff above is solved, we can think about librarizing
flashrom.
There are probably quite a few additional issues I didn't list above.
Comments are appreciated. Patches are welcome.

Regards,
Carl-Daniel

-- 
http://www.hailfinger.org/