On Tue, 20 Jan 2015 09:06:28 +0100 (CET), Daniel Stenberg wrote: > > It's easy enough to resolve this at the curl end, by casting the > > potentially > > errorneous calculation to a signed long (see attachment). > > I merged and pushed that one now.
Thanks. > > The second issue, which is the long-standing one, is that modern OS > > releases > > and C libraries don't need bsdsocket.library opening manually, as the C > > library can handle this. This means the #ifdef in amigaos.c/h needs to be > > revised. As this is now the odd case, I suggest adding a check to > > Configure > > for the amitcp/socketbasetags.h header, and only enabling that code if it > > is > > present. Note there is a select/WaitSelect definition in curl_setup.h, and > > an Amiga-specific in tool_operate.c, which also need to be removed (see > > attachments) if building using the C library functions. > > I'm not sure I follow here. Are you saying that the suggested patch should be > applied no matter what to build for Amiga? If you mean the attached one, no. That was just an example to point out which #ifdefs I was referring to. > When it comes to doing things to > configure and adjusting the build for Amiga I'm all for that but we really > need either an Amiga user to do that and verify that it gets right, or even > better an Amiga build added to the autobuilds system so that others can > verify > that the build goes through. Yes, agreed. I've only built here for AmigaOS 4 (which *requires* that code to be removed) and AmigaOS 3 against clib2 (which doesn't need it). Against libnix or ixemul or other c libraries it's likely that it might be required. It's a bit of a mish-mash which is why I suggested checking for amitcp/socketbasetags.h. Maybe something in config-amigaos.h to switch between the two build types might be better? The default can be left as-is, but it will avoid having to patch a number of files when the C library is providing the network interface. > My guess is that you find that many other fellow Amiga hackers working on > curl > these days... :-/ No, but the library certainly gets used. > > Lastly, in curl 7.40.0 there is some IPv6 specific code which is > > getting compiled in even with IPv6 disabled: > > I pushed a fix for that just now, it'd be great if you tested it. Clearly the > number of users building curl without ipv6 support nowadays is shrinking as > you're the first to report this! That's probably a good thing! The fix is working, thanks. Chris ------------------------------------------------------------------- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
