~dr|z3d
@RN
@T3s|4_
@eyedeekay
@orignal
@postman
@zzz
%Liorar
%acetone
%cumlord
%snex
+FreefallHeavens_
+Xeha
+ardu
+hk
+mareki2p
+onon_
+r00tobo
+radakayot
+uop23ip
AHON1
Arch
Danny
DeltaOreo
Irc2PGuest28644
Irc2PGuest4450
Irc2PGuest70966
Irc2PGuest87589
Meow
Onn4l7h
Onn4|7h
SigSegv
altonen
derivative
maylay
not_bob_afk
plap
poriori_
profetik1
qend-irc2p
shiver_
simprelay
solidx66
thetia
u5657
usr001
weko_
woodwose
zer0bitz
mareki2p
I did some constexpr programming original, should work with C++17, but my compiler defaulted to C++20. My work is here: github.com/mareki2p/i2pd/commits/i18n_ver1
orignal
and how did it improve?
StormyCloud
StormyCloud going dark for a few hours - cake.i2p/file/IgAMHtalWa_X7HjQFzTJ0wXp7MjAKEscc1ssfAHIj_09CwLlWIEk/s6.gif
zzz
good luck StormyCloud, thanks for your help
mareki2p
No more std::map<...> is being constructed. The idea is basically the same as of std::string -> std::string_view change. You have bunch of data (string_view to string_view mapping pairs) and you construct std::map from it. The map is a dynamic container capable of data being added and removed from it during run-time. But in this case we don't need the functionality, but we are paying for it anyways. The map nee
mareki2p
ds to be constructed at run-time, during program start-up. Basically reading bunch of read-only memory (string_views), calling malloc/new to allocate new tree nodes and constructing the tree (std::map is a tree). Onnly to be never modified during run-time again. What a waste. My approach massages the data to be in sorted order during compilation-time. And leaves the data as-is sitting in a std::array container.
mareki2p
Array is better than map because there are no pointer chasing, no need for memory allocation during run-time / start-up. The drawback is no ability to modify the array during run-time, but we don't need this functionality, so no drawback. Then the search needs to be adjusted, I'm using binary search (lower_bound) in an array instead of pointer chasing and cache missing of a tree/map.
mareki2p
In a nutshell: I build the search data structure at compile-time instead of at program start-up. I don't allocate memory. I utilitze the cache during lookup better. The compile time might be longer, but run-time is better. The idea is exactly the same as of the transition from std::string to std::string_view.
snex
normally when you try to hit a b32 via an http service tunnel, and there is no website there, the router returns a 504 gateway timeout. but every so often it returns a 403. what in the world is going on here? im not sure if this only happens in i2p+ or if it is in vanilla too
snex
tried to track down the code handling this but not able to find anything helpful through string searches
snex
its also saying 403 Connection Reset which is weird because 403 is the Forbidden code
dr|z3d
403 usually means "try again"
dr|z3d
it usually indicates a timeout of sorts. not always.
snex
but why is it doing that instead of the 504
snex
i also sometimes get a 502
dr|z3d
it's a different error.
snex
ok i need to know WHY it is doing this
dr|z3d
if the tunnel is running but not the service, you'll see a 5* class error. if there's some sort of timeout and server is running, you may see a 403.
snex
hm
dr|z3d
so, in brief, 403 usually means "try again".
snex
is it possible that whatever is running there is not an http server but some other kind of service?
dr|z3d
sure, that's also possible. a server can be instructed to send a 403.
snex
the server is not sending it. the router / service tunnel proxy is sending it
dr|z3d
if you're seeing the same 403 error consistently, then the likelihood is it's not serving http requests.
snex
its not consistent
dr|z3d
so, probably a timeout issue then.
snex
i dont understand why it would 403, that is not an appropriate response code
dr|z3d
403, as indicated in I2PTunnelHTTPClient.java, could mean one of a number of things.
dr|z3d
Access Denied, Bad Protocol, Bad URI, SSL rejected, Disabled..
snex
the 403 http status code means forbidden or access denied
snex
we shouldnt be using it as some kind of catch-all
snex
also this is what im seeing: "HTTP/1.1 403 Connection Reset" which is not even listed in that java file
snex
the string "403 Connection Reset" isnt anywhere in the code
dr|z3d
git grep -i "connection reset"
snex
yeah that doesnt give anything helpful
snex
installer/resources/proxy/reset-header.ht <- thats the file being served to me
dr|z3d
the client file I referenced has all the codes that are served, with backup html.
dr|z3d
private final static String ERR_REQUEST_DENIED = ...
snex
thats not what im getting
dr|z3d
well, I don't know what you're looking at, code-wise, or what you're trying to achieve.
snex
im simply trying to curl a b32. the response im getting is exactly whats in that .ht file i listed
dr|z3d
and? you're getting a 403. what happens if you try in a browser session?
snex
its inconsistent. sometimes it gives me the 504 as expected, but rarely it gives this 403
dr|z3d
try curl with the user agent "MYOB/6.66 (AN/ON)"
snex
that is the user agent being sent
dr|z3d
dunno, then. either a 403 or a 502 amounts to the same thing. "no".
snex
they do not though. a 403 could be returned by a genuine site on the other end
snex
i dont want my proxy telling me 403 on its own. thats absurd
snex
there are only 2 reasons a proxy should give me a 403: 1. the proxy requires auth and i didnt supply it. 2. the site on the other end generated the 403