WTF Bites
-
It's not an interpreter. It's a configuration file processor.
And, really, x86(_64)/ARM machine code just turns out to be a really convenient way to store the configuration data.
-
Soooo. What do you do if there's a problem with the font size regarding a particular information system?
-
Restarted/updated firefox, and get greeted by this. Firefox was happy just minutes ago, so wtf? Note that this is for gmail, and all google sites (incl. youtube) seem to return similar errors. Great.
Anyway, bing.com works, and apparently disabling spdy/http2 makes the error go away. Makes me think that it's not google who needs to fix their site...
Filed under: The browser developers will need to fix the browser first before you can visit this site.
-
-
-
@RaceProUK Huh. Ok, then. Guess it's google after all.
I sill think some more information and perhaps a "I-dont-care,-give-me-grumpy-cat"-button (like with the other cert problems) would have been helpful. Especially considering that they don't care about that issue enough to even warn about it outside of spdy/http2.
-
Huh. Ok, then. Guess it's google after all.
Well, GeoTrust, really, for using an insecure cert.
-
@RaceProUK Looks like Google provided the full chain. I'd be surprised if GeoTrust hadn't made the SHA256 cert available.
-
You'd maybe have to relink the entire thing, which can be painful enough.
Yeah, our game requires upwards of six gigabytes of RAM to link. Oh the joys of building with the engine source!
-
@Tsaukpaetra said in WTF Bites:
Yeah, our game requires upwards of six gigabytes of RAM to link. Oh the joys of building with the engine source!
There is that.
Then there is linking with LTO, which can draw out build times for even smallish projects to around 20 minutes. (It's even better when the LTO makes no measurable difference, but people keep insisting that it's vital for the speedz.)
-
I've been desperately trying to find out exactly how much caffeine is in Nescafé. So far with help from Google Scholar I've managed to find exactly TWO sources, one's a study from a Spanish magazine from 1991 (which says 3.9%) and one Kenyan study from 2010 which says 3.12%. Even a website dedicated to knowing how much caffeine is in coffee brands doesn't seem to have it (they do have a page for Apples though, in case you were wondering).
I thought it was like one of the biggest brands in the world?
-
@anonymous234 said in WTF Bites:
Even a website dedicated to knowing how much caffeine is in coffee brands doesn't seem to have it
Because it's not really coffee
-
@anonymous234
http://www.nescafeusa.com/coffee101/caffeine-productsNESCAFE Clasico
Regular approximately 65 mg
Decaf 2-5 mgNESCAFE Taster’s Choice:
House Blend approximately 65 mg
French Roast approximately 65 mg
100% Colombian approximately 65 mg
Flavored approximately 25-30 mg
Decaf House Blend 2-5 mg
Caffeine content based on typical range of coffee powder used in 6 oz cup.
-
@loopback0 That's useful, but why can't they tell you the actual quantity of powder they consider "one cup"? I'm sure they know the number.
By my own measurements, 2 small teaspoons = 2.4g = 84mg of caffeine, so to get 65mg means either there's less caffeine than I thought or their spoons are fucking tiny.
-
@anonymous234 Their "single serving" packets are ~1.4g, so seems like it's 1tsp.
-
@cvi
The issue isn't (completely) Firefox, it's (at least partly) Google Mail's certificate...The certificate validation path calls back to a SHA1-signed root CA, so my guess is that Firefox took the SHA1 certificate deprecation to be "we're going to not trust any certificate with a SHA-1 signature anywhere in the verification chain." Judging by Chrome Dev (57.0.2986.0 on my test box right now), Chrome is going to do the same thing as soon as the January stable version update drops within the next week.
Ooooooops.
-
It's a root certificate. It doesn't matter even if it's signed with MD2.
-
@izzion
And now it looks like they've pulled the site offline (presumably to work on the issue). Damn it, I should have made my previous screen shot a little better.
-
@PleegWat
Hm, yeah, and one of the certificates I maintain crosses back to a SHA-1 root and it's working in Chrome Dev. I don't have access to the original certificate chain I took that screenshot with, since GMail is down now, wonder what the problem was/is, then...
-
-
@TimeBandit
Heh, it's working on one ISP but not another. I wonder if their IP block got partially hijacked or one of these providers is doing some sort of shenanigans that broke with the SHA-1 deprecation...
-
so my guess is that Firefox took the SHA1 certificate deprecation to be "we're going to not trust any certificate with a SHA-1 signature anywhere in the verification chain."
Although apparently only if you try to talk SPDY or HTTP/2 with it. But yeah, seems like I got the developer firefox version too, so that might be related (not sure why, blame package manager etc. ... although I find the element inspector thingy useful on occasion).
-
-
@loopback0 said in WTF Bites:
@anonymous234 Their "single serving" packets are ~1.4g, so seems like it's 1tsp.
was it just me or did anyone else drop into this thread right here and get lost trying to convert how many gigabits of data are in 1 teaspoon?
-
@darkmatter said in WTF Bites:
@loopback0 said in WTF Bites:
@anonymous234 Their "single serving" packets are ~1.4g, so seems like it's 1tsp.
was it just me or did anyone else drop into this thread right here and get lost trying to convert how many gigabits of data are in 1 teaspoon?
That's if using 20nm tech. It gets better with newer stuff.
-
@darkmatter said in WTF Bites:
@loopback0 said in WTF Bites:
@anonymous234 Their "single serving" packets are ~1.4g, so seems like it's 1tsp.
was it just me or did anyone else drop into this thread right here and get lost trying to convert how many gigabits of data are in 1 teaspoon?
Depends on the teaspoon media, I suspect. What's the structural integrity of a fibre teaspoon?
-
@darkmatter said in WTF Bites:
@loopback0 said in WTF Bites:
@anonymous234 Their "single serving" packets are ~1.4g, so seems like it's 1tsp.
was it just me or did anyone else drop into this thread right here and get lost trying to convert how many gigabits of data are in 1 teaspoon?
Depends on the teaspoon media, I suspect. What's the structural integrity of a fibre teaspoon?
glass or carbon?
-
@darkmatter said in WTF Bites:
was it just me or did anyone else drop into this thread right here and get lost trying to convert how many gigabits of data are in 1 teaspoon?
Already solved
-
we have a process where we generate a csv file and it gets upload to a 3rd party (via web form I think, doesn't matter). the 3rd party is changing over to a webservice soon so we have to update.
the ws methods takes xml strings directly as parameters.
-
@coldandtired that's a helluva upload speed!
-
@Tsaukpaetra Link-time optimization
-
-
Link-time optimization
A technique that, while theoretically a great idea, usually wins almost nothing as it usually only optimises out cheap tests and at a cost of reducing code locality (due to the increased number of function replicas due to inlining). What's worse, LTO can be extremely expensive to perform on a large program.
The benefits are larger with some languages other than C and C++, where LTO can enable more extensive optimisations to be performed.
-
Link-time optimization
A technique that, while theoretically a great idea, usually wins almost nothing as it usually only optimises out cheap tests and at a cost of reducing code locality (due to the increased number of function replicas due to inlining). What's worse, LTO can be extremely expensive to perform on a large program.
The benefits are larger with some languages other than C and C++, where LTO can enable more extensive optimisations to be performed.
I wonder if such optimizations include literally copy pasting a const string five times into the binary despite being defined once?
-
@Tsaukpaetra said in WTF Bites:
I wonder if such optimizations include literally copy pasting a const string five times into the binary despite being defined once?
Not usually; those tend to be coalesced as they're in the DATA or RODATA sections, not the TEXT section.
-
@dkf LTO can be useful when you're dealing with callback heavy-code.
A concrete example would be an optimization code where the energy function was passed as a callback (virtual interface). In that particular example the energy function did fairly little work itself, but ended up called millions of times. LTO gave a very noticeable speed-up (IIRC it managed to inline the callback, after speculative devirtualization). Rewriting the code to enable proper inlining would have maybe improved things further, but in this case it wasn't worth the effort. ;-)
-
@coldandtired that's a helluva upload speed!
Considering the data is probably largely redundant, you'd get like 40MB in ... like 2 minutes? Colour me unimpressed.
-
2 minutes
Well hey, if you can maintain outflow for 2 minutes then more power to ya.
-
@coldandtired that's a helluva upload speed!
Considering the data is probably largely redundant, you'd get like 40MB in ... like 2 minutes? Colour me unimpressed.
~75 MB, more like; each sperm only carries half the data, usually with some errors.
-
2 minutes
Well hey, if you can maintain outflow for 2 minutes then more power to ya.
I assumed he meant that the latency was about 2 minutes. Hey, I guess it's still better than 1 minute!
-
@Tsaukpaetra said in WTF Bites:
I wonder if such optimizations include literally copy pasting a const string five times into the binary despite being defined once?
Not usually; those tend to be coalesced as they're in the DATA or RODATA sections, not the TEXT section.
Huh, must have been a mistake in the linker or something, the clean build didn't do it, but one that I was working on for a week messing with flags had it in there several times.
Maybe I did get it right the first time and a clean build would have shown it?
I just don't like waiting two hours for a full build just to see if something's missing...
-
LTO can be useful when you're dealing with callback heavy-code.
That's one of the few cases where it helps, but usually the gain isn't worth it as it also tends to stiffen the code by eliminating the separation between interfaces and implementations. If you're doing high-performance computing the costs can be worthwhile, especially when you've got a large main loop in your simulation engine or analysis code, but most programs are actually I/O-bound these days and so can tolerate a fair bit of inefficiency.
Curiously, I'm currently working with people doing (a novel kind of) HPC. :) But that's got even more esoteric rules on what's optimal…
-
@Tsaukpaetra said in WTF Bites:
Huh, must have been a mistake in the linker or something, the clean build didn't do it, but one that I was working on for a week messing with flags had it in there several times.
Did you have incremental linking enabled? I would guess that this could lead to having multiple copies of certain stuff after a few iterations. A clean build would start from scratch, so it would "fit" with your observations.
-
@Tsaukpaetra said in WTF Bites:
Huh, must have been a mistake in the linker or something, the clean build didn't do it, but one that I was working on for a week messing with flags had it in there several times.
Did you have incremental linking enabled? I would guess that this could lead to having multiple copies of certain stuff after a few iterations. A clean build would start from scratch, so it would "fit" with your observations.
Maybe?
I'm just focused on making sure the "server" parts get left out of the "client" builds.
It seems UE4 by default lets clients become servers at will, which is undesirable in our case as we obviously don't want them connecting to our DB ('cause passwords and all that) or even trying. 'course this "clients can be servers" thing also triggers Windows Firewall, which we're also trying to kill (because, seriously, client shouldn't be a server!)...
-
but usually the gain isn't worth it as it also tends to stiffen the code by eliminating the separation between interfaces and implementations.
Well, it's hardly worse than inlining the damn thing from the get go. :-)
But, yeah, the above example was in a HPC-y context, albeit a fairly simple one (single node, runtime in tens of minutes) and mostly for playing around. Which is why I didn't want to spend too much time on "fixing" the code (~templates instead of virtuals, so the callbacks become inlineable by default).
-
@Tsaukpaetra Hmm, yeah, haven't had a chance to play around with UE4.
Also,
@Tsaukpaetra said in WTF Bites:
connecting to our DB ('cause passwords and all that)
Hmm... wtf-sense is tingling. ;-)
-
@Tsaukpaetra Hmm, yeah, haven't had a chance to play around with UE4.
Also,
@Tsaukpaetra said in WTF Bites:
connecting to our DB ('cause passwords and all that)
Hmm... wtf-sense is tingling. ;-)
Oh, It's not that bad, I finished getting rid of the HTTP requests (that go out to our website and returns DB data) from it. ;)
-
@Tsaukpaetra said in WTF Bites:
@Tsaukpaetra Hmm, yeah, haven't had a chance to play around with UE4.
Also,
@Tsaukpaetra said in WTF Bites:
connecting to our DB ('cause passwords and all that)
Hmm... wtf-sense is tingling. ;-)
Oh, It's not that bad, I finished getting rid of the HTTP requests (that go out to our website and returns DB data) from it. ;)
I'm assuming you left the S off intentionally, too.
-
@Tsaukpaetra said in WTF Bites:
@Tsaukpaetra Hmm, yeah, haven't had a chance to play around with UE4.
Also,
@Tsaukpaetra said in WTF Bites:
connecting to our DB ('cause passwords and all that)
Hmm... wtf-sense is tingling. ;-)
Oh, It's not that bad, I finished getting rid of the HTTP requests (that go out to our website and returns DB data) from it. ;)
I'm assuming you left the S off intentionally, too.
Nope! We don't have an SSL certificate yet!
Edit: Today the CEO is supposed to be doing the verification thingies to get a validation or whatever so we can get a Code Signing cert (I'm still under the impression that I'm the only one who knows anything about this stuff, so I'm prodding them along).
-
Well, it's hardly worse than inlining the damn thing from the get go.
LTO is like that, but across independently-compiled library boundaries.