quic: QuicCID refactors and docs #286
Conversation
| // similar to what we can do with TraceEvents... that would | ||
| // allow us to pass the QuicCID directly to Debug and have it | ||
| // converted to hex only if the category is enabled so we can | ||
| // skip committing resources here. |
There was a problem hiding this comment.
Fully agree 👍
It’s a bit tricky because that means we can’t use fprintf() directly anymore, but that’s probably pretty okay actually given that having a C++-y replacement would also help solve issues like nodejs/node#28761.
|
Updated! |
jasnell
force-pushed
the
quiccid_refactors
branch
from
cf28446 to
74f6fb5
Compare
|
@addaleax ... added a new commit that avoids copying in QuicCID when not necessary to do so. Can you please take a look when you get a moment? |
|
@jasnell Did you benchmark this? (I can also do that if you like.) My best intuition would be that the new code actually adds non-trivial overhead without much gain. Copying an |
|
|
Much of the time, QuicCID is wrapping a `const ngtcp2_cid*`, when that is the case, avoid copying it's value and use the pointer directly. Otherwise, allocate an internal buffer.
jasnell
force-pushed
the
quiccid_refactors
branch
from
74f6fb5 to
f9ea38a
Compare
|
Ok, take another look and let me know if that's better :-) |
PR-URL: #286 Reviewed-By: Anna Henningsen <[email protected]>
PR-URL: #286 Reviewed-By: Anna Henningsen <[email protected]>
Much of the time, QuicCID is wrapping a `const ngtcp2_cid*`, when that is the case, avoid copying it's value and use the pointer directly. Otherwise, allocate an internal buffer. PR-URL: #286 Reviewed-By: Anna Henningsen <[email protected]>
|
Landed! |
PR-URL: nodejs#286 Reviewed-By: Anna Henningsen <[email protected]>
PR-URL: nodejs#286 Reviewed-By: Anna Henningsen <[email protected]>
Much of the time, QuicCID is wrapping a `const ngtcp2_cid*`, when that is the case, avoid copying it's value and use the pointer directly. Otherwise, allocate an internal buffer. PR-URL: nodejs#286 Reviewed-By: Anna Henningsen <[email protected]>
There are two commits here: