On Mon, Feb 04, 2008 at 07:08:46PM +0300, Dmitry Kurochkin wrote: > 2008/2/4, David Roundy <[EMAIL PROTECTED]>: > > On Sat, Feb 02, 2008 at 09:21:34PM -0700, [EMAIL PROTECTED] wrote: > > > Ran into the same problem Zooko did on MacOS X 10.4 (Tiger): the curl > > > library is older and doesn't support curl_multi_timeout. I believe > > > using a default timeout is a reasonable substitute. > > > > Thanks for the patch! It would be nicer in general to code the test using > > an explicit check for libcurl_multi_timeount (which insures us against > > either a future version that doesn't have this function, or an earlier > > version that does have it, in case you got the version number when it was > > added wrong), but I'm not leaning towards adding this check myself. I just > > mention this for future reference. AC_CHECK_LIB does this, its second > > argument is a function to try linking against. I see the pipelining check > > has the same issue... > > I do not think AC_CHECK_LIB will do for pipelining. First, there is no > function for pipelining, it is an enum. And besides, pipelining option > was introduced somewhere in 7.17 if I am not mistaking. But until 7.18 > it does not work as darcs curl module expects. So we really need to > check for curl version >= 7.18. In addition we can check if > CURLMOPT_PIPELINING enum is present. But I do not think it gives us > much benefit (and I do not know how to do it with autoconf).
Ah yes. That's precisely when a version check makes sense: if a feature is added before it works. Also you're right, checking for enums is much trickier than checking for functions. -- David Roundy Department of Physics Oregon State University _______________________________________________ darcs-devel mailing list darcs-devel@darcs.net http://lists.osuosl.org/mailman/listinfo/darcs-devel