This game is well loved by strategy enthusiasts, but could be equally fun for strategy wimps if it was more accessible. The game is unabashedly designed to be a struggle, and is so even on Easy if you haven't mastered all the strategy elements, while if you have, Medium is probably wehere you should be playing.
Essentially it needs a better scale-down in difficulty level. The best way to achieve this will probably be via AI tweaking.
You are promoting the pattern which produces the problem.
Your steps
1 - open filename.temp for writing 2 - write into filename.temp 3 - rename filename.temp to filename
The problem with this is that the operating sytsem has never been asked to push the data into the actual filename.temp, just to rename it over filename. This can result in a zero byte filename after a crash, as reported in the article.
Correct method:
1 - open filename.temp for writing 2 - write into filename.temp 3 - fsync filename.temp and check return value for errors 4 - close filename.temp 5 - rename filename.temp to filename
"BUT BUT BUT," everyone says "That's crazy, overriding a filename with an unwritten file who would want to do that?"
The fact is, sometimes that is what you would want. Not all renames are in the case where you are attempting to do an atomic update of file state. The posix system calls don't express this intent, so you have to do it yourself with use of fsync.
As others point out, there's a price here. fsync() will spin up the disk, when it isn't necesssary. Sometimes all you need is that you have the old data or the new data, and you don't care which, so fsync is heavier than you want. True, but your choices are heavier than you want, or risk no data at all.
Basically, many people have been assuming POSIX guarantees something it doesn't. Maybe we can get a new system call out of the hooplah that allows people to request this behavior when it is neeeded. In the meantime, fix the damn applications.
Sure, but there's pushing it down the road 50 years, and pushing it down the road until the expansion of the sun. I'm willing to settle for the second.
I've bicycled through the extensive wind farms in the Livermore hills many times, and they never seemed noisy to me. Maybe if you put one right in the middle of a city, it might be loud to people say living in the windmill building?
I'm going to guess that mostly it's just an outdated sense of aesthetics.
So when are they going to fix the obnoxious shit where firefox jumps on top of all my other windows when it finishes rendering a page?
Seriously you guys, this was in firefox 1.x and fixed in 2.x Regressing over a timescale of years is kind of mindboggling.
Re:looks like it still loses history
on
BASH 4.0 Released
·
· Score: 1
histappend is part of making bash act reasonably.
In the words of Aaron Crane:
# Append to the history file on shell exit, don't overwrite (so multiple
# exiting shells don't race to scribble over your saved history)
shopt -s histappend
# Keep lots of history around
export HISTSIZE=1000000 HISTFILESIZE=1000000
# Store history in a different file, so it won't get overwritten if you
# don't have these settings
export HISTFILE=~/.bash_history_safe
# Write unsaved history immediately before emitting each prompt
export PROMPT_COMMAND='history -a'
Amazingly, some replay code fails to play ogg vorbis files gaplessly. I have no idea what the transition from track to track looks like in their code, and don't want to.
I belive sexual readiness has a psychological component which does not necessarily correspond to the physiological component. Maybe your idea is reasonable, but I don't believe your current argument is sufficient.
Pornography unfortunately is essentially only legal by technicality, where by pornography I mean media which captures people who are having actual sex. Paying people to have sex is illegal, and is called prostitution. Pornography where you use a camera exists in a loophole where you claim not to be paying the people for sex, but they happen to have the sex for the time while you are paying them.
I'd like to see prostitution both legalized and placed under moderately strict requirements. We have requirements on many things such as consumption of alcohol, practicing law, and so on. I believe a reasonable set of restrictions can be crafted on prostitution that will create a better environment for sex workers than the one we have currently.
Yeah, and superdisks floppies were so important for most users.
SuSE had working modular kernels in 2.0.x. The only reason that you were picking kernels on some distributions in 2.0.x is because the whole initrd business hadn't been gotten beaten into place on most distributions. The modules loaded fine, it was just they hadn't figured out a standard way of accessing them when you couldn't get to the disk yet.
Sure, Linux got modular in the 1.x to 2.x series. In 2.0.x you really didn't need to recompile. Some people did. Some distributions hadn't figured it out yet. It wasn't a technical issue.
But really this is just an installation issue. Imagine you had to "recompile" the kernel for your disk driver, the distribution can handle this. It's solvable by tools.
Linux has been "finished" for a long long time, in that it works, it provides a posix environment, and it doesn't crash much. 1.2.13 was fine, really.
It's only if you care about support for particular hardware and performance in various scenarios that ongoing work is required.
Paging always has a cost right. Allocated and unused memory has a cost too. So if a page is allocated and never paged out, at some point the cost of not paging it out in the first place is larger than the cost of paging it out. THe question is just where that crossover point is.
Is it a minute, an hour a day?
I certainly have programs which run for days or months, and they sometimes have many many megabytes of unused memory.
If your box has more memory than will ever be used by all your os, apps, and files, then there isn't a problem, since a sane paging algorithm won't push that aside for anything, since there's no pressure.
Either your OS is completely insane, or the programs you are using are making use of more files and memory than you think they are.
You are clearly lying. The parse-time exceptions you get from broken indentation are blatantly clear, and logic flow changes you get from indenting incorrectly are only possible if you fail to understand the whole concept of indentation in code.
Your bloated fat programs like mozilla and your game will have plenty of pages they never plan to use, because of memory fragmentation and other issues. In a typical malloc implementation, you will end up with pages that have NO data in them but still can't be returned to the oprating system. Paging lets these be dumped at essentially no cost.
It's just a question of tuning to find the sweet spot. You definitely want SOME paging, but the ratio to memory has certainly been falling over the years.
Python is a good choice for 98% of programming tasks. Most languages are not easier nor mor more managable. Most are faster. Python has useful object oriented features.
You seem to be willing to comment on Python without knowing much about it. I guess I should learn to ignore ignorant commentary, you're right.
Slashdot Mirror
This game is well loved by strategy enthusiasts, but could be equally fun for strategy wimps if it was more accessible. The game is unabashedly designed to be a struggle, and is so even on Easy if you haven't mastered all the strategy elements, while if you have, Medium is probably wehere you should be playing.
Essentially it needs a better scale-down in difficulty level. The best way to achieve this will probably be via AI tweaking.
This problem is not specific to ext4. Portable code (such as KDE aims to be) *must* call fsync to employ this pattern correctly and safely.
The application has a bug. The end.
ext4 is being changed to try to minimize the problem, but it isn't the one with the bug.
You are promoting the pattern which produces the problem.
Your steps
1 - open filename.temp for writing
2 - write into filename.temp
3 - rename filename.temp to filename
The problem with this is that the operating sytsem has never been asked to push the data into the actual filename.temp, just to rename it over filename. This can result in a zero byte filename after a crash, as reported in the article.
Correct method:
1 - open filename.temp for writing
2 - write into filename.temp
3 - fsync filename.temp and check return value for errors
4 - close filename.temp
5 - rename filename.temp to filename
"BUT BUT BUT," everyone says "That's crazy, overriding a filename with an unwritten file who would want to do that?"
The fact is, sometimes that is what you would want. Not all renames are in the case where you are attempting to do an atomic update of file state. The posix system calls don't express this intent, so you have to do it yourself with use of fsync.
As others point out, there's a price here. fsync() will spin up the disk, when it isn't necesssary. Sometimes all you need is that you have the old data or the new data, and you don't care which, so fsync is heavier than you want. True, but your choices are heavier than you want, or risk no data at all.
Basically, many people have been assuming POSIX guarantees something it doesn't. Maybe we can get a new system call out of the hooplah that allows people to request this behavior when it is neeeded. In the meantime, fix the damn applications.
Oh, you've used Linux for over 15 years, eh? What distribution were you using in 1993?
Sure, but there's pushing it down the road 50 years, and pushing it down the road until the expansion of the sun. I'm willing to settle for the second.
I've bicycled through the extensive wind farms in the Livermore hills many times, and they never seemed noisy to me. Maybe if you put one right in the middle of a city, it might be loud to people say living in the windmill building?
I'm going to guess that mostly it's just an outdated sense of aesthetics.
You could also just set your filesystem to full async mode.. buhhh.
So when are they going to fix the obnoxious shit where firefox jumps on top of all my other windows when it finishes rendering a page?
Seriously you guys, this was in firefox 1.x and fixed in 2.x Regressing over a timescale of years is kind of mindboggling.
histappend is part of making bash act reasonably.
In the words of Aaron Crane:
# Append to the history file on shell exit, don't overwrite (so multiple
# exiting shells don't race to scribble over your saved history)
shopt -s histappend
# Keep lots of history around
export HISTSIZE=1000000 HISTFILESIZE=1000000
# Store history in a different file, so it won't get overwritten if you
# don't have these settings
export HISTFILE=~/.bash_history_safe
# Write unsaved history immediately before emitting each prompt
export PROMPT_COMMAND='history -a'
Yeah offtopic.. but..
Even windows thinks a newline is \n, it just stores them in files as \r\n but presents them to programs as \n.
Yes, it's psychotic.
This dicsussion was over some time ago.
Uh, what are "nearly root" permissions?
Unix doesn't support this.
Amazingly, some replay code fails to play ogg vorbis files gaplessly. I have no idea what the transition from track to track looks like in their code, and don't want to.
KDE isn't a single program, so describing it succintly is harder.
THE END.
I belive sexual readiness has a psychological component which does not necessarily correspond to the physiological component. Maybe your idea is reasonable, but I don't believe your current argument is sufficient.
Pornography unfortunately is essentially only legal by technicality, where by pornography I mean media which captures people who are having actual sex. Paying people to have sex is illegal, and is called prostitution. Pornography where you use a camera exists in a loophole where you claim not to be paying the people for sex, but they happen to have the sex for the time while you are paying them.
I'd like to see prostitution both legalized and placed under moderately strict requirements. We have requirements on many things such as consumption of alcohol, practicing law, and so on. I believe a reasonable set of restrictions can be crafted on prostitution that will create a better environment for sex workers than the one we have currently.
It's in the mould.
Taxing travel depresses travel. Taxing fuel depresses inefficient travel more than efficient travel.
Now which seems to make more sense?
Yeah, and superdisks floppies were so important for most users.
SuSE had working modular kernels in 2.0.x. The only reason that you were picking kernels on some distributions in 2.0.x is because the whole initrd business hadn't been gotten beaten into place on most distributions. The modules loaded fine, it was just they hadn't figured out a standard way of accessing them when you couldn't get to the disk yet.
Sure, Linux got modular in the 1.x to 2.x series. In 2.0.x you really didn't need to recompile. Some people did. Some distributions hadn't figured it out yet. It wasn't a technical issue.
But really this is just an installation issue. Imagine you had to "recompile" the kernel for your disk driver, the distribution can handle this. It's solvable by tools.
Linux has been "finished" for a long long time, in that it works, it provides a posix environment, and it doesn't crash much. 1.2.13 was fine, really.
It's only if you care about support for particular hardware and performance in various scenarios that ongoing work is required.
because len(foo) and foo.len() are soooooooooooo different.
They both make use of a common protocol. Only the purist need care.
Paging always has a cost right. Allocated and unused memory has a cost too. So if a page is allocated and never paged out, at some point the cost of not paging it out in the first place is larger than the cost of paging it out. THe question is just where that crossover point is.
Is it a minute, an hour a day?
I certainly have programs which run for days or months, and they sometimes have many many megabytes of unused memory.
If your box has more memory than will ever be used by all your os, apps, and files, then there isn't a problem, since a sane paging algorithm won't push that aside for anything, since there's no pressure.
Either your OS is completely insane, or the programs you are using are making use of more files and memory than you think they are.
Uh, son. Notice where I said 'reformat'. Maybe you should, you know, look that up.
You are clearly lying. The parse-time exceptions you get from broken indentation are blatantly clear, and logic flow changes you get from indenting incorrectly are only possible if you fail to understand the whole concept of indentation in code.
If your editor sucks so much that it can't reformat without using tabs, then you need to move into the nineties, son.
You still want paging.
Your bloated fat programs like mozilla and your game will have plenty of pages they never plan to use, because of memory fragmentation and other issues. In a typical malloc implementation, you will end up with pages that have NO data in them but still can't be returned to the oprating system. Paging lets these be dumped at essentially no cost.
It's just a question of tuning to find the sweet spot. You definitely want SOME paging, but the ratio to memory has certainly been falling over the years.
You misunderstood.
Python is a good choice for 98% of programming tasks. Most languages are not easier nor mor more managable. Most are faster. Python has useful object oriented features.
You seem to be willing to comment on Python without knowing much about it. I guess I should learn to ignore ignorant commentary, you're right.