Re: [Powertop] [RFC][PATCH 1/2] Adding report generator facility

Hi Sergey, Sergey Senozhatsky wrote: Done. Please check v2 patchset (after list moderator will approve those big letters). The null formatter is needed only because the code that uses the report facility is called even when the reports are not needed. The function one_measurement() (which calls to other reporting functions) first called from make_report() with 1 second parameter. And that report will be thrown away. In the comments there is a note that this is to warm-up the system. I don't know whether it is really needed. After that make_report() produces real reports inside the loop by calling the one_measurement() function. Also one_measurement() is called from calibration code, and that code doesn't need any reports too. So we need to decide when to call reporting functions from one_measurement(). We could check the report_type there. But what type of report maker should we create when we don't need the reports? HTML? CSV? Sounds no good. Providing a flag inside the report maker whether we need to produce the reports and check it upon entry to every method is no good too. We could create report maker and provide it to the needed functions by pointer. But what to provide when we need not report? NULL? Then we have to check for it in every function. Lot's of unnecessary checks. We could combine two mentioned approaches -- create report maker only when we actually need to produce the report, and call report creating functions from one_measurement() with non-NULL pointer to report (and pass it down to all other functions that need it). Then we could get rid of null report-maker. But now it is easier way to deal with such usage of report facility. What you think? It is very hard to refactor this code. -- Best regards, Igor Zhbanov, Expert Software Engineer, phone: +7 (495) 797 25 00 ext 3806 e-mail: i.zhbanov(a)samsung.com ASWG, Moscow R&D center, Samsung Electronics 12 Dvintsev street, building 1 127018, Moscow, Russian Federation