On 3/3/2023 3:08 AM, Stuart Henderson wrote:
> On 2023/03/02 22:41, Brian Callahan wrote:
>> Hi Volker --
>>
>> On 2/27/2023 1:18 PM, Volker Schlecht wrote:
>>>
>>>
>>> On 2/27/23 18:49, Stuart Henderson wrote:
>>>
>>>> Ah, looking at that file note the difference between "curl" and
>>>> "libcurl", it seems that it tries to use libcurl first and then there
>>>> are fallbacks to use wget/curl binaries. So it seems the lines you
>>>> patched wouldn't have been reached in the first place if libcurl
>>>> detection was working properly?
>>>
>>> It's a fallback yes, but the fallback is implemented in function
>>> posixMethod(), which takes the fallback as a parameter, but only uses it
>>> if haveLibcurl==false:
>>>
>>> posixMethod <- function(utility) {
>>> if (haveLibcurl)
>>> "libcurl"
>>> else if (nzchar(Sys.which(utility)))
>>> utility
>>> else
>>> ""
>>> }
>>>
>>> I'm quite certain that libcurl detection works and that
>>> haveLibcurl==true on OpenBSD, but unless the OS is one of "Windows",
>>> "Darwin" or "Linux", posixMethod() is never called, libcurl is never
>>> detected, and an empty string is returned...
>>>
>>
>> Is there an updated patch on this?
>
> I've looked at the surrounding code now and while I don't read R very
> well, I think the existing diff is ok (R itself already depends on curl
> so no new dep). Seems a bit strange that they'd go out of their way to
> avoid trying to autodetect libcurl except on named OS but *shrug*
>> ~Brian
Thanks Stuart. I'll test and commit this evening.
~Brian
