On Mon, Jan 11, 2021 at 2:35 AM Daniel Stenberg wrote:
>
> On Mon, 11 Jan 2021, Jeffrey Walton via curl-library wrote:
>
> > $ lsb_release -a
> > Distributor ID:Ubuntu
> > Description:Ubuntu 20.04.1 LTS
> > Release:20.04
> > Codename:focal
> >
> > $ command -v wget
> >
On Mon, 11 Jan 2021, Jeffrey Walton via curl-library wrote:
$ lsb_release -a
Distributor ID:Ubuntu
Description:Ubuntu 20.04.1 LTS
Release:20.04
Codename:focal
$ command -v wget
/usr/bin/wget
$ wget -O cacert.pem 'https://curl.haxx.se/ca/cacert.pem'
Unable to locally verify
Hi Everyone/Daniel,
$ lsb_release -a
Distributor ID:Ubuntu
Description:Ubuntu 20.04.1 LTS
Release:20.04
Codename:focal
$ command -v wget
/usr/bin/wget
$ wget -O cacert.pem 'https://curl.haxx.se/ca/cacert.pem'
--2021-01-11 02:09:02-- https://curl.haxx.se/ca/cacert.pem
Resolving
On 1/10/2021 5:33 PM, Daniel Stenberg via curl-library wrote:
On Sun, 10 Jan 2021, Fabian Keil via curl-library wrote:
The attached patch seems to work around the problem.
Busy-looping probably isn't the right solution, though.
Yeah, that's just not the right fix.
When the code returns
On Sun, 10 Jan 2021, Patrick Monnerat wrote:
Not for me with NSS/3.60. Test 300 works fine here.
Fedora 33 stock nss 3.59.0 here.
Fired up a VM with an older distro. Current curl master with NSS/3.26.2 - also
ran fine. (except that unit test 1621 fails to build there)
Not dimissing your
On 1/10/21 10:57 PM, Daniel Stenberg wrote:
On Sun, 10 Jan 2021, Patrick Monnerat wrote:
Plase note that nss seems cursed (even in the master branch): test
suite gives valgrind errors and test 300 never finishes.
Not for me with NSS/3.60. Test 300 works fine here.
Fedora 33 stock nss
On Sun, 10 Jan 2021, Fabian Keil via curl-library wrote:
The attached patch seems to work around the problem.
Busy-looping probably isn't the right solution, though.
Yeah, that's just not the right fix.
When the code returns there (http_proxy.c line 346) due to CURLE_AGAIN having
been
On Sun, 10 Jan 2021, Midnight Wonderer wrote:
Thanks for explaining!
This is potentially affecting most HTTP-based services in existence.
But client retry is not practical, mostly because there is no standard for
it. Some clients took, like, a minute to try the next IP address. Take curl
as
On Sun, 10 Jan 2021, Patrick Monnerat wrote:
Plase note that nss seems cursed (even in the master branch): test suite
gives valgrind errors and test 300 never finishes.
Not for me with NSS/3.60. Test 300 works fine here.
--
/ daniel.haxx.se
| Commercial curl support up to 24x7 is
Fabian Keil via curl-library wrote on 2021-01-09:
> Fabian Keil via curl-library wrote on 2021-01-08:
>
> > using a curl binary built from git master 060658f634, test 1061
> > works reliably for me when valgrind is being used but is flaky
> > without it:
[...]
> The problem seems to be that
On 1/10/21 5:59 PM, Daniel Stenberg wrote:
On Sun, 10 Jan 2021, Patrick Monnerat wrote:
Could be turned at first as:
void abcd(struct Curl_easy *data)
abcd(conn->data);
Sure it can, and if you can come up with a way where that helps then
good for you. In my head that's just an
On Sun, 10 Jan 2021, Patrick Monnerat wrote:
Could be turned at first as:
void abcd(struct Curl_easy *data)
abcd(conn->data);
Sure it can, and if you can come up with a way where that helps then good for
you. In my head that's just an intermediate step that I don't think is
necessary to
On Sun, 10 Jan 2021 at 20:44, Daniel Stenberg wrote:
> I would rather we focued on what exactly the problem and use case you have
> with the existing algoritm is that makes you not able to just perhaps extend
> the timeout a little?
Good point!
To add some background:
As we all know, if you
On 1/10/21 2:32 PM, Daniel Stenberg wrote:
On Sun, 10 Jan 2021, Patrick Monnerat via curl-library wrote:
Considering the large number of references, I would suggest starting
to fix this "bottom up" by smaller commits rather than having a big
patch, even if it increases the reference count at
On Sun, 10 Jan 2021, Midnight Wonderer wrote:
Realistically speaking, anyone who put 20 addresses for a hostname doesn't
expect the client to try all 20 addresses;
This can be made an option with a sane default. (like 4, or some number
acquirable via statistic mean, or simply turn off)
It
On Sun, 10 Jan 2021, Patrick Monnerat via curl-library wrote:
Considering the large number of references, I would suggest starting to fix
this "bottom up" by smaller commits rather than having a big patch, even if
it increases the reference count at first while the work is not complete.
On 1/8/21 3:51 PM, Daniel Stenberg via curl-library wrote:
Hi all,
Here's something that's been itching in me for a long time. Something
I've even (twice!) tried to fix. Both times my patch ended up so huge
and complicated for me that I had to give it and postpone the idea.
Put shortly:
17 matches
Mail list logo
