Gerald Murray wrote: >> Ren? Rebe wrote: >>> I take a look into what is necessary to get SANE into the LSB over >>> the weekend. >>> >> Did you already find out something? >> > Hi Till, > I took a quick look. It looks like the lsb project wants headers > --stripped of all comments and programmer remarks; > the library would be empty functions, as 'basic' as possible. > > Possibly headers: sane.h, saneopts.h would be enough, as > lsb tools would only test for the presence of sane, so none of > the libraries in /usr/lib/sane/* would be tested? > /usr/lib/libsane.so would be the basic library. > > Does the lsb project expect some testing code to be written? > What tests would be needed, or expected, if any?
What we need is two interfaces of SANE, once the frontend/library interface, so that one can have scanning functionality in LSB-packaged applications, and second, the library/backend interface to allow distribution-independent LSB-based scanner driver packages. The LSB expects also testing code. Usually a program which exercises all the interfaces which are selected as requirement for the LSB (ideally, one should require all interfaces of the library). A test program could look like this: http://svn.easysw.com/public/cups/trunk/cups/testcups.c This is for CUPS but one could do a similar thing for SANE. The test program should not depend on a particular scanner model. One could auto-detect a scanner and if none is there use the test backend. The individual drivers in /usr/lib/sane/ do not need to be tested. Important are only the interfaces for new drivers and for applications with scanning functionality. Till