> On Mon, Feb 23, 2009 at 11:58 PM, Alex Rousskov > <[email protected]> wrote: >> On 02/23/2009 10:52 AM, Kinkie wrote: >> >>>> * Declare when needed, for example inside for(): >>>> Â size_type j; >>>> Â for (j=0;j<len_;j++) { >>> >>> Do you mean that this should read >>> for (size_type j=0;j<len_;++j) >>> ? >> >> Yes, I do. IIRC, there are many other examples where variables are >> declared too early. This is a minor flaw, with no real effect in most >> cases, but it is a good habit to develop and becomes important when you >> deal with non-integral variables. >> >>> Next steps, mempooling SBufStore and its backing storage. >> >> MemPooling is needed but does not affect the design or most of the >> implementation code. There is no serious rush to implement this, IMO. > > Already done in rev 9366. Had to add a second bool flag to MemBlob, > but that's fine. > >> It looks like all the big items in the previews review are closed. Do >> you want me to do another review soon or should I wait for some other >> milestone/event? > > Any time you want and have the time to spare, I leave it to your > judgement. > There's some pending cleanup: the current TODO list on top of my head is: > - rename SBufStore to MemBlob and give it its new home > - rename SBuf to Buffer (I don't remember if we had agreed on a > different name, there were quite a few that flew around) and rename > files accordingly > - rename testSBuf files and classes. > - check that the function names in debugs() statements match their > actual name (there's been quite a lot of shuffling)
Please use debugs(..., HERE << "blah") for that sort of thing at levels >1. I've also just added MYNAME for prettier but identical use on the top two levels. They can save a lot of tracing confusion. > - define a few more MemPool String sizes and match them with those in > estimateCapacity > - maybe shuffle function implementations around SBuf.cc(i) to match > their definition order in SBuf.h > Amos
