Age | Commit message (Collapse) | Author |
|
|
|
The version with the vector, that can be disabled etc, is “very” slow, so we
use unique_ptr when we don’t need to disable it, and when it only contains
one function
|
|
“It is possible to select(2) or poll(2) for completion by selecting the
socket for writing”
Yeah, sure, “writing”…
|
|
|
|
|
|
duplication
|
|
See
https://lab.louiz.org/louiz/biboumi/issues/3215
https://github.com/andikleen/glibc/blob/master/time/strptime_l.c#L746-L747
for why strptime() sucks
We use std::get_time now, to parse the date and time. And we parse the
timezone by hand.
fix #3215
|
|
|
|
fix coverity CID 134470
|
|
fix coverity CID 134469
|
|
fix #3069
|
|
fix #3207
|
|
|
|
Because some plateform accept Z and z, but some only accept z…
|
|
ref #2948
|
|
Thank you, e2e tests
|
|
And make things a little bit clearer
|
|
Avoids a conflict between /usr/include/uuid.h and
/usr/local/include/uuid/uuid.h on freebsd
|
|
|
|
If anybody knows why fedora accepts both, but it only works with %z on
debian, please tell me.
|
|
|
|
|
|
|
|
|
|
At the moment, result-set-management is not implemented, the whole history
(well, at most 1024 messages) is returned.
|
|
|
|
“stable”
Use strftime instead
|
|
|
|
|
|
|
|
|
|
Can’t remember why I did this, but that must be a stupid reason. Everything
must work even with a timeout of 0.
|
|
Because c-ares may close one of its socket, even if it’s not the one that
has the event. Otherwise we may not know when a socket has been removed from
our poller (automatically, when close()ed) and this leads to bugs.
|
|
This reverts commit 5328d0806fdc5becb9344b4d4320787a2b7c0712.
|
|
|
|
|
|
|
|
|
|
Do not use a singleton for the database.
fix #3203
|
|
It’s $CURRENT_YEAR
|
|
|
|
|
|
ref #3067
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|