Dear all,
Thanks for your time so-far.
I send a mail to the libcurl development list and got a helpful reaction [1]! I
changed my patch accordingly. We now don't have to check for a CURLE_WRITE_ERROR
anymore and curl_easy_perform() will return CURLE_OK (0) on success.
Please review!
commit c5588e94f5f0e66636b16a4ee26f0dbcf48b44c9
Author: Boris Schrijver <bo...@pcextreme.nl>
Date: Wed Dec 9 23:32:36 2015 +0100
qemu-img / curl: When fetching Content-Size use GET instead of HEAD.
A server can respond different to both methods, or can block one of the two.
Signed-off-by: Boris Schrijver <bo...@pcextreme.nl>
diff --git a/block/curl.c b/block/curl.c
index 8994182..1677d0c 100644
--- a/block/curl.c
+++ b/block/curl.c
@@ -598,6 +598,7 @@ static int curl_open(BlockDriverState *bs, QDict *options,
int flags,
curl_easy_setopt(state->curl, CURLOPT_HEADERFUNCTION,
curl_header_cb);
curl_easy_setopt(state->curl, CURLOPT_HEADERDATA, s);
+ curl_easy_setopt(state->curl, CURLOPT_CUSTOMREQUEST, "GET");
if (curl_easy_perform(state->curl))
goto out;
curl_easy_getinfo(state->curl, CURLINFO_CONTENT_LENGTH_DOWNLOAD, &d);
[1] http://curl.haxx.se/mail/lib-2015-12/0041.html
--
Met vriendelijke groet / Kind regards,
Boris Schrijver
PCextreme B.V.
http://www.pcextreme.nl/contact
Tel direct: +31 (0) 118 700 215
> On December 8, 2015 at 8:56 PM Michael Tokarev <m...@tls.msk.ru> wrote:
>
>
> 08.12.2015 00:23, Boris Schrijver wrote:
> []
> > It's is therefore better to use only the GET request method, and discard the
> > body at the first call.
>
> Nooooooo! Please NOOOO!
>
> Fetching a large file once might be too long already.
> Fetching it twice is twice as long. Oh well!..
>
> Thanks,
>
> /mjt
