|
Janin posted:21759446
|
# ? Jun 11, 2011 01:00 |
|
|
# ? May 15, 2024 16:21 |
|
If you use numbered changelists the revision numbers can go up a lot faster than the number of commits. Our p4 repo at work is on rev 27826 with only 12671 submitted changelists.
|
# ? Jun 11, 2011 01:25 |
|
http://www.infoq.com/presentations/Development-at-Google Not really that great of a talk but goddamn that poo poo is insane. All of google's code is in one repo! I wish I had the time to improve our little 20 dev HG based system.
|
# ? Jun 11, 2011 03:58 |
|
Plorkyeran posted:If you use numbered changelists the revision numbers can go up a lot faster than the number of commits. Our p4 repo at work is on rev 27826 with only 12671 submitted changelists. Yeah, p4 gets kind of goofy with that. Pretty much any operation that COULD feasibly end with a submit gets a changeset number.
|
# ? Jun 11, 2011 05:14 |
|
When I worked in vidya games it got pretty crazy sometimes. I checked an easter egg into the non-gold fork (e.g. the one we used for loving around) which changed one AI script and one object. It ended up pulling over 300 binaries into the changelist.
|
# ? Jun 11, 2011 06:42 |
|
Scaramouche posted:When I worked in vidya games it got pretty crazy sometimes. I checked an easter egg into the non-gold fork (e.g. the one we used for loving around) which changed one AI script and one object. It ended up pulling over 300 binaries into the changelist. Yeah but that was still only one changelist. Where I work the engineering depot for all projects are held in common with just different branches and we have like 200k changes, (Over the past 7 years) the assets for each game are in their own depot and can run up to 100k changes per game.
|
# ? Jun 11, 2011 14:35 |
|
NotShadowStar posted:No loving way. Pics or it didn't happen. At my last job, we had one SVN repo for every single project in the history of company; all 200+ of them. You'd do an initial commit on a new project, and it would be #96345. The next day, you'd commit your first change, and it would be #100032. A year ago, a bunch of the devs were pushing for moving to Mercurial or git, but it got vetoed because, I poo poo you not, the Director of Development was upset that "we can't have every project in the same repo, so no way."
|
# ? Jun 11, 2011 16:58 |
|
MOOMIN.EXE posted:one of the developers at my jorb used to insist on doing seperate commits on every single file he modified I do that. It makes git bisect more useful than it already is.
|
# ? Jun 11, 2011 17:03 |
|
BonzoESC posted:I do that. It makes git bisect more useful than it already is. I think you are confused.
|
# ? Jun 11, 2011 17:36 |
|
SavageMessiah posted:http://www.infoq.com/presentations/Development-at-Google Isn't Apache similar? I seem to recall massive numbers browsing one project. HTTPD is at 1134699, http://svn.apache.org/repos/asf/httpd/
|
# ? Jun 11, 2011 18:32 |
|
ColdPie posted:I think you are confused.
Yes, broken code with failing tests gets committed. This is why you run tests before building and deploying. At the same time, I can go back to the reasoning for any line of code, the test that explains the reasoning, and see the commit comment for why I did it. Comments get stored in the git history which can be blamed to individual lines and disappear automatically when the lines they refer to are changed. In-line comments can't do that. If something gets broken, I can run git bisect against the particular test that requires that feature until I see where it breaks. I don't worry about repo bloat because commits are really loving cheap in git.
|
# ? Jun 11, 2011 19:18 |
|
BonzoESC posted:
For us, we usually squash the commits before we push to remote. Part of the reason is that we're in transition from Perforce to git, and right now we have a git layer that commits to Perforce on the backend. The other reason is that we don't need all those commits, we need a few (or single) buildable and deployable commits, that can be reverted to a stable state. This will probably change as more of our systems move into continuous build and deployment mode. Of course this is after all the unit tests and integration tests pass as well as getting a few code reviews. I am just happy we're getting rid of Perforce.
|
# ? Jun 11, 2011 20:40 |
|
BonzoESC posted:Yes, broken code with failing tests gets committed. This is why you run tests before building and deploying. What you described is different from committing each file separately, which makes bisecting basically useless and brings you back to the CVS-era of non-atomic commits.
|
# ? Jun 11, 2011 23:00 |
|
Janin posted:Well it's not like I'm by myself; plus, we use Perforce, not subversion. I think the current change rate is 20 commits per minute. Yeah, I cheated too, since we're on Perforce as well.
|
# ? Jun 13, 2011 20:22 |
|
We have to submit powerpoint slides each week for a group meeting. For my next slide I added a bullet to the Issues section: "Overcoming the cognitive dissonance I get from seeing yards and meters on the same graph." I do so love working here.
|
# ? Jun 13, 2011 21:46 |
|
I just found this gemcode:
|
# ? Jun 13, 2011 23:56 |
|
JediGandalf posted:I just found this gem Don't worry, I validated the input!
|
# ? Jun 14, 2011 00:04 |
|
Lexical Unit posted:We have to submit powerpoint slides each week for a group meeting. For my next slide I added a bullet to the Issues section: "Overcoming the cognitive dissonance I get from seeing yards and meters on the same graph." You forgot to mention that neither axis is labeled.
|
# ? Jun 14, 2011 00:47 |
|
JediGandalf posted:I just found this gem Wait, maybe my demented, drunken brain isn't working, but isn't CheckField a function that makes sure all characters in a string are members of that third argument, and while == true is probably redundant, and OK = true makes me hurt inside for some reason, and there is no reason for nesting (in what is probably a function which runs whenever an input textbox thing loses focus) it still works correctly? Hell breaking out CheckField into another function is something that would merit a commendation compared to the code I run into at work.
|
# ? Jun 14, 2011 03:53 |
|
Geekner posted:Don't worry, I validated the input! Worst part is the "return OK" at the end, I'm ok with nested if statements to some extent (I've seen some horrible horrible nested if's in my time)
|
# ? Jun 14, 2011 04:12 |
|
aux posted:Worst part is the "return OK" at the end, I'm ok with nested if statements to some extent (I've seen some horrible horrible nested if's in my time) I like the implication that the variable OK could be set to false, and "return OK" could therefore be a very bad thing.
|
# ? Jun 14, 2011 09:17 |
|
qntm posted:I like the implication that the variable OK could be set to false, and "return OK" could therefore be a very bad thing. Would you prefer "isOK"? It really doesn't seem like a big deal to me. A boolean called "OK" can really only mean one thing, and "if (OK)" reads naturally even of OK is false.
|
# ? Jun 14, 2011 09:21 |
|
A boolean called "OK", like "TRUE", "FALSE", "SUCCESS" and "FAILURE", looks like a constant. Especially when you start using it in conditionals. "if(OK)" reads like "if(true)" or "if(1)". "if(OK == false)" just looks like nonsense. "isOK" would be an improvement, but "isValid" would be better.
|
# ? Jun 14, 2011 12:48 |
|
Just return, OK?!
|
# ? Jun 14, 2011 13:37 |
|
Welp. Every time you think you've seen the most incompetent data security on the internet, there comes along a new and exciting fuckup in basic security. Hackers steal 200,000 credit card numbers by changing the URL after logging into a different account. The NYT article this page links to is even more hilarious. quote:In the Citi breach, the data thieves were able to penetrate the bank’s defenses by first logging on to the site reserved for its credit card customers. jonjonaug fucked around with this message at 22:27 on Jun 14, 2011 |
# ? Jun 14, 2011 22:18 |
|
quote:One security expert familiar with the investigation wondered how the hackers could have known to breach security by focusing on the vulnerability in the browser. “It would have been hard to prepare for this type of vulnerability,” he said. The security expert insisted on anonymity because the inquiry was at an early stage. What? I suspect he insisted on anonymity because he's just smart enough to realize he's dumb. I'm not a security expert at all, but have heard of these type of attacks for a decade.
|
# ? Jun 14, 2011 22:22 |
|
Well, he's right, you know. It really is pretty hard to predict that users are going to just go and edit the URL. I mean, it's got tons of numbers and letters in it. How will they know what the right ones are?
|
# ? Jun 14, 2011 23:47 |
|
code:
|
# ? Jun 15, 2011 00:47 |
|
Wow, all that and they mix malloc and delete.jonjonaug posted:Welp. Every time you think you've seen the most incompetent data security on the internet, there comes along a new and exciting fuckup in basic security.
|
# ? Jun 15, 2011 01:58 |
|
The inline is a nice touch. Clearly this piece is performance-critical.
|
# ? Jun 15, 2011 02:00 |
|
US banks think that asking you for both your password and your security questions is two-factor authentication.
|
# ? Jun 15, 2011 02:01 |
|
pseudorandom name posted:US banks think that asking you for both your password and your security questions is two-factor authentication. But how could anyone know what year my father was born??
|
# ? Jun 15, 2011 14:08 |
|
Pardot posted:
This is genius, and flawed.
|
# ? Jun 15, 2011 14:11 |
|
Pardot posted:
I'm calling bull. Clearly all the clock_t's should be volatile.
|
# ? Jun 15, 2011 15:02 |
|
I'm not a C++ programmer, what the gently caress does that even do?
|
# ? Jun 15, 2011 15:04 |
|
He tries to find out the total memory size by continuously asking the system for a fresh chunk of memory, until the system tells him to gently caress off by returning a null pointer. Alternatively, it also terminates by some magically defined time out, probably for those fabled 'virtual memory' system the new kids on the block have.
|
# ? Jun 15, 2011 15:10 |
|
It also uses memset to force the system to actually allocate the memory pages rather than just promising to (as most modern systems do when you call malloc).
|
# ? Jun 15, 2011 15:44 |
|
What is "Slot1"? You might think that it's just the first cell of a 3x8 table. No, it is in fact the only cell of a 1x1 table that is inside a 1x2 table, that is inside a 3x2 table. The three cells under Slot1 are in a 1x3 table. The text in some, but not all, cells is in a div just for kicks. And every table contains a shitload of inline styles. Like '<td class="fart_butt" style="font-size:8pt; border-bottom:0">'
|
# ? Jun 15, 2011 16:10 |
|
My dad's on the board of a credit union and he told me they were looking at some software package for online banking that would require users to type in their password four times for every login attempt and use some sort of keystroke frequency analysis to determine whether the typist was the original user. This was back in 06, I think. The software they ended up with (may or may not be the same system, I don't know) will crash with the standard ASP exception page if it get unexpected input in some places - like when I gave it my out of state address :\Wheany posted:
You found one of my current employer's sites, I guess?
|
# ? Jun 15, 2011 16:33 |
|
|
# ? May 15, 2024 16:21 |
|
Hughlander posted:I'm calling bull. Clearly all the clock_t's should be volatile. http://www.stanford.edu/group/wonglab/jiangh/seqmap/ Grad student code
|
# ? Jun 15, 2011 17:35 |