On 04/03/2019 11:08, Adriano dos Santos Fernandes wrote: > On 04/03/2019 10:57, Alex Peshkoff via Firebird-devel wrote: >> On 3/4/19 3:15 PM, Vlad Khorsun wrote: >>> 04.03.2019 14:01, Alex Peshkoff via Firebird-devel wrote: >>>> On 3/4/19 2:57 PM, Vlad Khorsun wrote: >>>>> I have additional question: is it true that fbclient now depends >>>>> on ICU ? >>>> As fas as I know - not. At least IntlManager.cpp is placed in >>>> src/jrd, which is used only in engine. >>>> Client is using other (OS-specific) tools for charsets conversions. >>> Look at UtilInterface::decodeTimeTz() - it calls >>> >>> TimeZoneUtil::decodeTime() >>> TimeZoneUtil::decodeTimeStamp() >>> Jrd::UnicodeUtil::getConversionICU() >>> >> Yes. Now I see - any ICU version was always loadable by some functions >> in common library but that functions were never called in fbclient. Now >> with TZ support it does... >> But I'm unsure how this can be fixed. Should we force all TZ support to >> be moved to server? May be taht makes sense - in a case when we start to >> work with TZ-related data we should already have connection with any >> engine. >> > I suppose that any "real" dependency here is to transform TZ-id to > string and vice-versa. > Let me correct myself. It's not to convert IDs to/from strings, but to convert UTC time to region time as string form, as in descriptors the data is stored in UTC.
It's an optional dependency in the sense that only when these new IUtil function are called ICU is needed. I do not think an utility function that converts time from regions should require a server call. Wire compress requires zlib in the client when used, it1s the same. I do not think optional (not required for previously functionality) ICU in the client is the end of the world or is worst than requiring server call to convert a time. Adriano Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
