Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> writes: > The way it was being done before was intentional & discused on list. > > See my original > https://lore.kernel.org/git/patch-v3-7.7-93a2775d0ee-20210730T092843Z-avarab@xxxxxxxxx/ > which did it pretty much like that, and Junio's subsequent > follow-up. I.e. this breadcrumb trail: > https://lore.kernel.org/git/?q=CURL_SOCKOPT_OK > >> -#if LIBCURL_VERSION_NUM < 0x071505 >> -#define CURL_SOCKOPT_OK 0 >> +#if LIBCURL_VERSION_NUM >= 0x071505 >> +#define GIT_CURL_HAVE_CURL_SOCKOPT_OK 1 >> #endif > > IOW we should drop this. I think that depends on the worldview. In a world in which [PATCH 2/2] is a good idea, i.e. "we have a comprehensive catalog of available cURL features, but it expresses its knowledge in one particular way, i.e. HAVE_X", the above, together with the change at the only use site in http.c, are very sensible changes. Given that we do not want to have too many conditionally compiled codepath, I certainly understand that the current approach to keep an ad-hoc list of features we care about may be your preference. I am not sure if that is viable longer term, though. I still am not decided. Thanks.
