Rich Megginson wrote: > The reason this works is because > * sizeof(long) == sizeof(PRFileDesc *) on both 32bit and 64bit > platforms > * We replace the sockbuf I/O functions with the PR I/O functions > before the lber functions have a chance to access this value
thats it, it came back now, thats why i made that change. > In fact, upon further investigation, it appears that we don't need any > of the compat_io stuff anymore. I can't find any trace of it in Fedora > DS, except in the start_tls_extop.c teardown code, and I don't see how > that can possibly work. In every other place, it has been commented out > or removed entirely. So I propose we just get rid of that code too - > basically, any reference to nslberi_io_fns. Does Sun DS still need/use it? unfortunately yes it does. its still using that hack to pass PRFileDesc around, just like your code did before changes made there "by Cheston". i dont personally think its a big deal to make related changes on Sun DS side so if you are determined to change LBER_SOCKET to native eg int you should probably go ahead with that if you really have to since we cant really be relying on the same old hack forever, maybe its time to finally fix it. if/when you do please send out a quick email with ci details or/and diffs so i can add relevant comments there and pass it to responsible individuals on our side. its been long forgotten i think & some folks, over here at least, might be relying on that code without realizing its actually a hack in llber that we kept alive for too long. _______________________________________________ dev-tech-ldap mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-ldap
