Alexey Galakhov wrote: > 2012/5/17 Sascha Hauer <s.hauer@xxxxxxxxxxxxxx>: > > Still you convert two different functions to a common name. Once again: > > Please keep s3c24xx_get_memory_size and add a s5p_get_memory_size > > function for the s5p SoC. > > It turned out to be useful when functions (or defines) have a spcific > > SoC name in them. This way you always know in which context a function > > is valid. Also it makes it possible to compile in all (in this case > > memory setup) functions in a single binary. > > I know that we do not follow this rule very strictly in barebox, but I > > won't accept patches that change places that do it right already. > > Ok. Sorry. > > BTW, there are functions like s3c_get_pclk(), and they are much worse > than get_memory_size regarding their portability. Newer S3Cs have > multiple clock domains, so there is more than one PCLK (i.e., > MSYS-PCLK and HSYS-PCLK). These functions are declared publilc, not > static, in a header file. They all are used in S3C24x0-specific code > only. Should they be renamed like s3c24xx_get_pclk() ? Should some of > them become static? Mostly a matter of taste. But sometimes these functions should use a common name: when they are used by a shared driver. If you call a S3C2440 related function by a S3C2440 related driver or board file, we should use a SoC specific name. When we call a function from a driver used for S3C2440 *and* S3C6410 it should use a non SoC specific name. This is a "should" and I fear my S3C24xx code is not perfect in this way. Regards, Juergen -- Pengutronix e.K. | Juergen Beisert | Linux Solutions for Science and Industry | http://www.pengutronix.de/ | _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
