info
Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
NAO undergoing a bit of unplanned maintenance
302 views
Skip to first unread message
dtype
Aug 11, 2014, 7:37:19 PM8/11/14
to
We were having some filesystem write errors, so I ended up rebooting NAO. Since fs hasn't been checked in awhile, it is fscking. As the structure of NAO is that there is a huge huge filesystem, this may take several hours.
At some point I'm upgrading the instance type NAO is running on. This may become a catalyst. In any case, working on getting things back up for now.
-drew
At some point I'm upgrading the instance type NAO is running on. This may become a catalyst. In any case, working on getting things back up for now.
-drew
yhelothar
Aug 12, 2014, 4:18:45 AM8/12/14
to
dtype
Aug 12, 2014, 12:45:27 PM8/12/14
to
On Tuesday, August 12, 2014 1:18:45 AM UTC-7, yhelothar wrote:
> We still don't have permission to view dumplogs and ttyrecs, and games aren't being reported on http://alt.org/nethack/gamesday.php today. Is the data for today's games lost? Also, http://alt.org/nethack/mostrecent.php hasn't been working for several months, can you fix that?
Thanks for letting me know. I'll look into these later today.
> We still don't have permission to view dumplogs and ttyrecs, and games aren't being reported on http://alt.org/nethack/gamesday.php today. Is the data for today's games lost? Also, http://alt.org/nethack/mostrecent.php hasn't been working for several months, can you fix that?
The data for the games should all still be good, and any games in progress should resume normally as long as nethack picks up the save/restore correctly.
The current issues are website specific, I think.
-drew
yhelothar
Aug 13, 2014, 3:25:14 AM8/13/14
to
Dumplogs and ttyrecs are showing up but just about everything else in http://alt.org/nethack/plr.php isn't updating. Top deaths and wishes haven't been updating for a while too (are those done monthly or something?)
The website has loads of issues, not just with slow updates but also with limited reporting features and search functionality. http://alt.org/nethack/gamesday.php only reports the first 100 games per day, which is absurd. Scummers like programluke and the88kod fill up the first 100 spots with #quits in only a few hours. Game data for the rest of the day is completely unsearchable.
Why is it that you can't search for all games on the entire site (dumplogs/ttyrecs) by date, [minimum] score, death reason, [max/current] dungeon level, etc? You can only lookup games for a specific player, by typing in the player's name and scrolling through their chronological list of games. Aside from that, the only thing searchable are top records, like the top 2000 scores, 100 fastest acensions, etc.
Why not let us search for site-wide dumplog txt data by keywords? e.g. "142 species extinct". Or at least allow search engines to crawl the dumplog directories so we can google it. You can't search for an extinctionist game without already knowing the player's name. Even if you know their name, you need to click on every one of their dumplogs to find the one you're looking for.
Why is it that you can't view anyone's 'All games - in raw logfile' except your own (after logging in)?
The website has loads of issues, not just with slow updates but also with limited reporting features and search functionality. http://alt.org/nethack/gamesday.php only reports the first 100 games per day, which is absurd. Scummers like programluke and the88kod fill up the first 100 spots with #quits in only a few hours. Game data for the rest of the day is completely unsearchable.
Why is it that you can't search for all games on the entire site (dumplogs/ttyrecs) by date, [minimum] score, death reason, [max/current] dungeon level, etc? You can only lookup games for a specific player, by typing in the player's name and scrolling through their chronological list of games. Aside from that, the only thing searchable are top records, like the top 2000 scores, 100 fastest acensions, etc.
Why not let us search for site-wide dumplog txt data by keywords? e.g. "142 species extinct". Or at least allow search engines to crawl the dumplog directories so we can google it. You can't search for an extinctionist game without already knowing the player's name. Even if you know their name, you need to click on every one of their dumplogs to find the one you're looking for.
Why is it that you can't view anyone's 'All games - in raw logfile' except your own (after logging in)?
Janis Papanagnou
Aug 13, 2014, 4:23:27 AM8/13/14
to
Am 12.08.2014 02:37, schrieb dtype:
> We were having some filesystem write errors, so I ended up rebooting
> NAO. Since fs hasn't been checked in awhile, it is fscking. As the
> structure of NAO is that there is a huge huge filesystem, this may
> take several hours.
Thanks for the information, and of course thanks for running the
> We were having some filesystem write errors, so I ended up rebooting
> NAO. Since fs hasn't been checked in awhile, it is fscking. As the
> structure of NAO is that there is a huge huge filesystem, this may
> take several hours.
site and for all your maintenance efforts!
>
> At some point I'm upgrading the instance type NAO is running on. This
> may become a catalyst. In any case, working on getting things back up
> for now.
can run any game without dying?[*] This is so fundamental a flaw
in many respects that it would be worth considering to fix it.[**]
Janis
[*] I've mentioned this in another recent thread about current
abuses of NAO.
[**] Yet I haven't inspected the source to see whether that fix
is easy to accomplish. Any ideas about that?
yhelothar
Aug 13, 2014, 5:22:05 AM8/13/14
to
^that would require an overhaul of the code that handles monster attacks, HP loss, movement points, and game restores, which hasn't been modified in decades and therefore would probably introduce new bugs and exploits. It could also hurt players who occasionally get legitimate hangups. It's a lot of risk and effort to fix something which hardly anyone is abusing. Besides there will always be workarounds.
Janis Papanagnou
Aug 13, 2014, 5:45:45 AM8/13/14
to
Am 13.08.2014 11:23, schrieb Janis Papanagnou:
>
> One question; do you see a chance to fix the flaw where everyone
> can run any game without dying?[*] This is so fundamental a flaw
> in many respects that it would be worth considering to fix it.[**]
Just had a brief look into the source. Everything necessary to fix
>
> One question; do you see a chance to fix the flaw where everyone
> can run any game without dying?[*] This is so fundamental a flaw
> in many respects that it would be worth considering to fix it.[**]
the basic flaw seems to be located in a single function hitmu(),
where a rearrangement of message display code and HP calculation
*might* fix the worst. Not quite sure about any interaction with
NAO's recovery mechanism. Any insights and concrete ideas welcome.
I'm on the fly now, so I'm inspecting the issue further later...
Janis
yhelothar
Aug 13, 2014, 9:01:23 AM8/13/14
to
Oh wow, is it really that easy? You should e-mail your solution to the devteam 20 years ago, when they realized the problem but decided it wasn't worth the trouble to fix.
Rearranging message display code will hardly 'fix the worst'. It would put a band-aid on 1% of the problem, and still require significant rewriting of mhitu.c and bug testing. And it certainly wouldn't stop a master of hangup exploits like myself, because there are plenty of tricks for outmaneuvering monsters.
It won't do a thing for monsters with multiple messages in their attacks, because that would cause bizarre message timing. I don't think people are OK with seeing "The goblin zaps a runed wand!--More--" and noticing their HP have already dropped to 0, even before pressing spacebar.
It won't work for monsters with multiple attacks per move. I don't think people are OK with seeing "The Archon wields a long sword named Sunsword!--More--" and noticing their HP have already dropped to 0.
It won't help with monsters that get multiple moves per turn. It won't help if you're surrounded and get attacked by 8 different monsters per turn. A hangup and restore will always stop monsters from attacking and give you a free move, which is more than enough to write elbereth or run away. You can't fix that with a little message display rearrangement. Fixing all the above problems without causing poor message timing or accidentally punishing legitimate hangups is a major programming effort and requires lengthy bug testing.
Rearranging message display code will hardly 'fix the worst'. It would put a band-aid on 1% of the problem, and still require significant rewriting of mhitu.c and bug testing. And it certainly wouldn't stop a master of hangup exploits like myself, because there are plenty of tricks for outmaneuvering monsters.
It won't do a thing for monsters with multiple messages in their attacks, because that would cause bizarre message timing. I don't think people are OK with seeing "The goblin zaps a runed wand!--More--" and noticing their HP have already dropped to 0, even before pressing spacebar.
It won't work for monsters with multiple attacks per move. I don't think people are OK with seeing "The Archon wields a long sword named Sunsword!--More--" and noticing their HP have already dropped to 0.
It won't help with monsters that get multiple moves per turn. It won't help if you're surrounded and get attacked by 8 different monsters per turn. A hangup and restore will always stop monsters from attacking and give you a free move, which is more than enough to write elbereth or run away. You can't fix that with a little message display rearrangement. Fixing all the above problems without causing poor message timing or accidentally punishing legitimate hangups is a major programming effort and requires lengthy bug testing.
jim in austin
Aug 13, 2014, 7:41:39 PM8/13/14
to
On Monday, August 11, 2014 6:37:19 PM UTC-5, dtype wrote:
> We were having some filesystem write errors, so I ended up rebooting NAO.
> Since fs hasn't been checked in awhile, it is fscking. As the structure
> of NAO is that there is a huge huge filesystem, this may take several hours
>
> We were having some filesystem write errors, so I ended up rebooting NAO.
> Since fs hasn't been checked in awhile, it is fscking. As the structure
> of NAO is that there is a huge huge filesystem, this may take several hours
>
> At some point I'm upgrading the instance type NAO is running on. This may
> become a catalyst. In any case, working on getting things back up for now.
Whatever you did a huge wad of nhqdb posts just appeared in my rss feed...
> become a catalyst. In any case, working on getting things back up for now.
ais523
Aug 14, 2014, 9:48:30 AM8/14/14
to
Janis Papanagnou wrote:
> One question; do you see a chance to fix the flaw where everyone
> can run any game without dying?[*] This is so fundamental a flaw
> in many respects that it would be worth considering to fix it.[**]
For what it's worth, this is fixed in NetHack 4, although it required
> One question; do you see a chance to fix the flaw where everyone
> can run any game without dying?[*] This is so fundamental a flaw
> in many respects that it would be worth considering to fix it.[**]
a massive amount of code restructuring. If you hangup in response to a
mosnter attacking you, the game will remember the partial turn so far,
and replay it when the game is reloaded. So you end up dying anyway.
The change is not at all trivial to apply to existing code,
unfortunately.
--
ais523
NetHack4 developer
Pat Rankin
Aug 19, 2014, 1:00:50 PM8/19/14
to
In nethack's development version, which may or may not
ever see the light of day, hangup handling was substantially
revamped several years ago. There was some minor reordering
(in dorecover(), I think) to try to prevent a hangup during
restore from clobbering that game. There's also a much bigger
change of swapping in a dummy windowing system that suppresses
further attempts at screen output and behaves as if the user
pressed ESC if/when the core asks for further input. Biggest
change is a compile-time option to use deferred hangup handling
(and I think it is enabled by default, but am not sure) such
that when a hangup signal is detected, it will just set a flag
instead of attempting to create a save file, and that flag gets
checked in moveloop() and possibly a couple of other places,
so hangup during monsters' movement will silently complete all
their movement before saving can take place.
I don't recall offhand what happens if the hero dies in the
interval between hangup and the code wending its way back to the
main loop. It might produce a file that yields "not healthy
enough to restore" during subsequent restoration or it might move
on to 'gave over' instead of saving. And if the latter, I'm not
sure whether the score and log files end up being updated for
that game. (Now that I think about it, it is almost certainly
the 'game over' path because dying doesn't make it back to
moveloop(); still not sure about record/logfile updating though.)
No doubt it introduces an assortment of new bugs, but
deliberately abusing hangup to cheat will become at least a bit
harder. There is definitely potential for surprises, like being
hit by a quantum mechanic or having ice you're standing on be
melted by a fire breather or wand zapper during the period when
there's no feedback displayed. In theory the dummy windowing
system could buffer message output and attach that to the
eventual save file, but I don't think that's very likely to get
implemented. It's quite a bit of extra complexity for something
that's not relevant to normal game play.
ever see the light of day, hangup handling was substantially
revamped several years ago. There was some minor reordering
(in dorecover(), I think) to try to prevent a hangup during
restore from clobbering that game. There's also a much bigger
change of swapping in a dummy windowing system that suppresses
further attempts at screen output and behaves as if the user
pressed ESC if/when the core asks for further input. Biggest
change is a compile-time option to use deferred hangup handling
(and I think it is enabled by default, but am not sure) such
that when a hangup signal is detected, it will just set a flag
instead of attempting to create a save file, and that flag gets
checked in moveloop() and possibly a couple of other places,
so hangup during monsters' movement will silently complete all
their movement before saving can take place.
I don't recall offhand what happens if the hero dies in the
interval between hangup and the code wending its way back to the
main loop. It might produce a file that yields "not healthy
enough to restore" during subsequent restoration or it might move
on to 'gave over' instead of saving. And if the latter, I'm not
sure whether the score and log files end up being updated for
that game. (Now that I think about it, it is almost certainly
the 'game over' path because dying doesn't make it back to
moveloop(); still not sure about record/logfile updating though.)
No doubt it introduces an assortment of new bugs, but
deliberately abusing hangup to cheat will become at least a bit
harder. There is definitely potential for surprises, like being
hit by a quantum mechanic or having ice you're standing on be
melted by a fire breather or wand zapper during the period when
there's no feedback displayed. In theory the dummy windowing
system could buffer message output and attach that to the
eventual save file, but I don't think that's very likely to get
implemented. It's quite a bit of extra complexity for something
that's not relevant to normal game play.
Janis Papanagnou
Aug 19, 2014, 6:14:17 PM8/19/14
to
On 19.08.2014 19:00, Pat Rankin wrote:
> In nethack's development version, which may or may not
> ever see the light of day [...]
> In nethack's development version, which may or may not
"All hope abandon ye who enter here."
Janis
code...@gmail.com
Aug 19, 2014, 6:52:16 PM8/19/14
to
On Tuesday, August 19, 2014 10:00:50 AM UTC-7, Pat Rankin wrote:
> In nethack's development version, which may or may not
> ever see the light of day, hangup handling was substantially
> revamped several years ago.
Aw, c'mon, guys, share what you've got. We miss you! We love you!
> In nethack's development version, which may or may not
> ever see the light of day, hangup handling was substantially
> revamped several years ago.
Jonathan Ellis
Aug 19, 2014, 8:40:51 PM8/19/14
to
<code...@gmail.com> wrote in message
news:e4391843-faf3-4af9...@googlegroups.com...
more than one main-version number jump...
-- JLE
Jorgen Grahn
Aug 20, 2014, 7:17:55 AM8/20/14
to
with my time ...
/Jorgen
--
// Jorgen Grahn <grahn@ Oo o. . .
\X/ snipabacken.se> O o .
Jorgen Grahn
Aug 20, 2014, 10:01:24 AM8/20/14
to
On Wed, 2014-08-20, Jorgen Grahn wrote:
> On Tue, 2014-08-19, code...@gmail.com wrote:
>> On Tuesday, August 19, 2014 10:00:50 AM UTC-7, Pat Rankin wrote:
>>> In nethack's development version, which may or may not
>>> ever see the light of day, hangup handling was substantially
>>> revamped several years ago.
>>
>> Aw, c'mon, guys, share what you've got. We miss you! We love you!
>
> I love you, although I sometimes wonder what else I could have done
> with my time ...
Speaking of too much Nethack: the other night I dreamt I saw a
> On Tue, 2014-08-19, code...@gmail.com wrote:
>> On Tuesday, August 19, 2014 10:00:50 AM UTC-7, Pat Rankin wrote:
>>> In nethack's development version, which may or may not
>>> ever see the light of day, hangup handling was substantially
>>> revamped several years ago.
>>
>> Aw, c'mon, guys, share what you've got. We miss you! We love you!
>
> I love you, although I sometimes wonder what else I could have done
> with my time ...
disenchanter. It was standing in a park a rainy night, and saw it
some distance away. It was of course a bright blue 'R'.
Janis Papanagnou
Aug 20, 2014, 10:57:01 AM8/20/14
to
On 20.08.2014 16:01, Jorgen Grahn wrote:
>
> Speaking of too much Nethack: the other night I dreamt I saw a
> disenchanter. It was standing in a park a rainy night, and saw it
> some distance away. It was of course a bright blue 'R'.
In a park, hmm.. - could have been part of an advertising panel (say,
>
> Speaking of too much Nethack: the other night I dreamt I saw a
> disenchanter. It was standing in a park a rainy night, and saw it
> some distance away. It was of course a bright blue 'R'.
"Read more books!", or so), but letters partly hidden by the plants.
Frankly, I'm not sure what annoys me more, a disenchanter in the park
that will just fail to reduce AC from my rotten +0 jeans and T-shirt,
or a park full of advertisement (as so typical a factor nowadays).
Janis
Jorgen Grahn
Aug 20, 2014, 11:51:01 AM8/20/14
to
On Wed, 2014-08-20, Janis Papanagnou wrote:
> On 20.08.2014 16:01, Jorgen Grahn wrote:
>>
>> Speaking of too much Nethack: the other night I dreamt I saw a
>> disenchanter. It was standing in a park a rainy night, and saw it
>> some distance away. It was of course a bright blue 'R'.
>
> In a park, hmm.. - could have been part of an advertising panel (say,
> "Read more books!", or so), but letters partly hidden by the plants.
No, it was standing by itself in the grass, and glowing. It looked
> On 20.08.2014 16:01, Jorgen Grahn wrote:
>>
>> Speaking of too much Nethack: the other night I dreamt I saw a
>> disenchanter. It was standing in a park a rainy night, and saw it
>> some distance away. It was of course a bright blue 'R'.
>
> In a park, hmm.. - could have been part of an advertising panel (say,
> "Read more books!", or so), but letters partly hidden by the plants.
like a letter from a neon sign, except you could clearly tell that it
would attack if you came closer.
I met it by accident. What I /really/ was doing was trying to watch
the HBO series "Girls" by peeking through the windows of random
strangers, hoping one of them was watching. And I don't follow that
series in real life.
> Frankly, I'm not sure what annoys me more, a disenchanter in the park
> that will just fail to reduce AC from my rotten +0 jeans and T-shirt,
> or a park full of advertisement (as so typical a factor nowadays).
Not around here, except maybe for a short period before the elections
in mid-September. Buses, trams and bus stops are full of them, though.
0 new messages