« first day (393 days earlier)      last day (2101 days later) » 

1:00 AM
@Adám It works if I put an echo between cat .code.tio and echo :endnamespace.
Should I make the change?
 
@Dennis Sure, go ahead. It removes the residual TIO_xxx namespace.
 
@Adám And that won't break anything? (I have no idea what that means.)
 
@Dennis The effect is almost identical to the current version. Currently, we generate a named namespace populated by Code, then copy everything into the root to allow running from Input. This version just creates an unnamed namespace and copies into root. Thus, root gets populated as before, but the random name doesn't pollute the workspace (and the wrapper cannot be detected from inside APL).
 
That sounds good.
@cairdcoinheringaahing Syncing now. MY also works now btw.
 
(think of an APL namespace as a JSON object, though it can contain programs/functions as well as data.)
You can "dot" into an APL namespace, so mynamespace.var gets the value of the var member of mynamespace. So with a bit of trickery, one can currently figure out what the intermediary namespace is called (TIO...). With this, one can't, as it is gone right away.
@Dennis Oh, you made the change already. Neat. Thanks.
 
1:11 AM
Yes, just did.
I think I understand the difference now. Thank you for explaining. :)
 
@Dennis I could detect that by looking at the list of available names as seen from Input. Now it is empty.
@Dennis So now, ⎕FIX establishes an unnamed namespace (:namespace without name). This namespace only exists as long as there exists a reference to it. The result of ⎕FIX is such a ref. ⎕NS then does a deep copy of that namespace into the root (#), but now there are no more references to the new namespace, so it is discarded.
 
Right.
 
1:44 AM
@Pavel Fedora isn't supported, and the list of dependencies sounds like I don't want to build it from source, especially since they mention that GCC 6 is incompatible with Hack.
 
2:09 AM
@Dennis The same binaries should theoretically work on any distro, right?
 
Only if all dependencies are fulfilled. And HHVM has a lot of dependencies.
 
Welp.
@Dennis There are instructions for building on Fedora 20: github.com/facebook/hhvm/wiki/….
 
Yes, I saw. I assume these instructions are for whatever version of Hack was current in 2015.
 
._. I'm starting to see the appeal of Windows now
 
Compiling from source wouldn't be easier on Windows.
 
2:23 AM
Well, in this particular case there aren't, but there are almost always prebuilt binaries for windows.
And since everyone has the same Windows, satisfying dependencies is easier, too.
 
Ha
No
Unless it's C#, compiling for Windows and managing dependencies is hell
 
Right. So you don't compile for Windows.
You use the prebuilt binary.
And there is a prebuilt binary, in 99% of cases.
 
and who do you think had to make that
 
Someone who is better at compiling things than I am.
 
Building binaries makes me sad.
 
2:26 AM
I didn't say I see the appeal of developing for Windows
 
I'm so used to being hand held, I can just write some code and it runs on everything.
 
furthermore Facebook has its own package manager for a lot of its C++ stuff, I highly doubt the "prebuilt binaries" are statically linked
Literally everything Facebook makes is a pain in the ass to install and takes like a gig of memory
 
@Pavel everyone has the same Windows is not true. There are three major consumer versions that come with different feature set, and a few (most likely rather different) server versions.
 
@Pavel Some people still use windows.
 
And if "compatibility mode" has taught me anything, it's that different Windows versions can be quite different when they want to.
 
2:28 AM
@Dennis For the most part, a Binary on one will work out of the box when copied to another. Problems only start with x86 and x86_64 differences.
 
@Dennis Well I've never seen software that didn't work because I have the wrong Windows, not counting decades-old software.
 
I have.
I've also seen software that stopped working because I uninstalled IE.
Or wouldn't install at all.
 
oh hell yes
 
Well, I guess I've just never tried uninstalling IE.
 
Because, you know, you absolutely need IE to display the EULA.
 
2:30 AM
Turns out everything and their dog uses Trident for rendering
When you take IE out that goes bye bye
 
OK, downloading the Hack source code created a 2.3 GB directory.
 
try building Proxygen from source sometime if you want some fun
I don't know what Facebook does to their C++
Or why they only have prebuilt binaries for Windows for a server side scripting language??
 
> HHVM is primarily tested on GCC 4.9 and 5. Using other versions may produce unexpected results, or may not even build at all.
Splendid, those are ancient.
 
LOL
and yet they pride themselves on their modern C++
all right I'll stop bashing
Just a warning though if HHVM tries to compile Folly or anything else there's a pretty good chance it'll eat a ton of RAM
I've been burned multiple times on their code which is why I'm kind of ranting
 
I'm fully aware that the chance this will actually result in a usable Hack interpreter are minuscule.
 
2:37 AM
pessimism is the best defense here
 
I mean, I'm trying to compile 2.3 GB of source code with a single core, 1GiB of physical RAM, and 1.9 GB of free disk space. What could possibly go wrong?
@ATaco Btw did you see this?
Sep 14 at 16:47, by Dennis
@ATaco Could you remove the second argument to fs.readFileSync in line 5 of MaybeLater's tokenizer.js? readFileSync doesn't take a callback, and this creates issues with the LTS version of Node I'm trying to switch to.
 
@ATaco See >:( your very unconventional, very unusual programming style has come back to bite you in the butt
 
Oh, I'll patch that now.
@Dennis Patched
 
2:52 AM
Works great. Thanks!
 
@ATaco did you also fix everything else that was wrong with the MaybeLater source :P
 
>:(
 
3:12 AM
And it went sideways.
/opt/hack/dev/hhvm/third-party/webscalesqlclient/src/mysys_ssl/my_aes_openssl.cc:121:18: error: aggregate ‘EVP_CIPHER_CTX ctx’ has incomplete type and cannot be defined
   EVP_CIPHER_CTX ctx;
                  ^~~
 
3:53 AM
@Dennis Could you please pull Crayon when you get a chance? (stupid enter key right net to my Shift key)
 
@Dennis that's a weird error
Quick Google suggests that's caused by a change in OpenSSL
 
Yes, a change in openSSL 1.1.0.
 
so you'll need to backport openssl 1.0.2 because compatibility is hard
That'll probably screw up your existing OpenSSL most likely
 
Which only leaves building openSSL 1.0.2 from source.
@ETHproductions Syncing.
 
@Dennis Thanks, seems to be working fine.
 
4:12 AM
@Dennis Your dedication to adding support for languages that practically no one uses, no matter how difficult, is trully amazing.
 
TIOBE disagrees with the practically no one uses part, but thanks. :)
 
Huh, really? I sorta figured someone would have requested it by now if it was used that much.
 
It's number 43, making it more popular than Julia, ML, Rust, and PowerShell.
 
I think all of those are more commonly used on PPCG than Hack, by virtue of me seeing all of those on PPCG except for Hack.
 
Maybe all Hack users are Facebook employees. ¯\_(ツ)_/¯
Jun 22 at 0:24, by Dennis
Yes, the tarball isn't working for me. TIO has openSSL 1.0.2.
Oh look, it would might have worked on Fedora 25.
 
4:29 AM
I could have sworn that's the current Fedora
 
Not the first time we had this conversation, I think.
 
Yep
 
Apr 21 at 16:43, by Pavel
Wait, 26? I installed Fedora last month, and got Fedora 23. ._.
 
Ok, wth. The fedora website is showing for me in Russian. I might have assumed this to be a weird one-time thing, but just yesterday I noticed hackmd also mysteriously switched to Russian for no reason at all.
 
eBay switched to Spanish for me, and started to displaying US items with Guaraní prices.
Also:
in The Nineteenth Byte, Sep 2 at 3:44, by Dennis
Re: racism, if you open godaddy.com from South America, they display all prices in Mexican Pesos. sighs
 
 
2 hours later…
6:19 AM
I seem to have jumped the openssl hurdle.
 
6:58 AM
@Dennis I got used that I get prices in Australian Dollars
no need to say, I don't use Australian dollars.
 
7:17 AM
@AndrewSavinykh :| wait what why would anyone do that
 
8:01 AM
@ASCII-only because for some people can't tell new zealand from australia
 
 
6 hours later…
1:46 PM
@Dennis the issue where the program doesn't terminate returned!
 
 
2 hours later…
3:29 PM
@ASCII-only <Kannadian Face Here>
 
@EriktheOutgolfer One of the arenas wasn't feeling well.
 
3:44 PM
@Pavel why do I know exactly what you mean
 
4:35 PM
@Dennis is it ill? was it ill but now it's fine? :p
 
It's all better now.
 
4:54 PM
good
 
 
1 hour later…
6:12 PM
@Dennis could you pull 2Deorstv please? Thanks in advance!
Also, could you change it so that it uses an SBCS please?
 
6:25 PM
For the byte count? Sure, as soon as I get home.
 
@Dennis Thanks!
 
7:26 PM
@Dennis actually, could you take 2Deorstv off of TIO? No answers have been posted using it yet, and only 2 links have be generated. It really needs to undergo a rehaul, and I don't want people using it between now and when it's finished.
 
@Dennis Just FTR, I've been running three of these pretty much since that post. No issues so far, though I use them as build servers (clang needs RAM), not web servers.
 
@cairdcoinheringaahing Aww, good luck fixing those. Can't wait to play around with its features after you finish.
 
 
1 hour later…
8:59 PM
@mınxomaτ I'll keep that in mind, thanks.
@cairdcoinheringaahing I don't like doing this, but since it's a recent addition and nobody seems to rely on it yet, OK.
 
@Dennis Thanks. I didn't want to bug you with a massive amounts of pull requests so I thought this would be better.
 
It's gone.
 
Bye bye 2Deorstv o/ :P
 
@Dennis Have you had an attempt at installing Curry using PAKCS as the compiler?
 
@Potato44 iirc yes
yesterday, by Dennis
Comes pre-built afaict, but I can't get it to work.
Seems to indicate he tried but failed.
 
9:11 PM
@Pavel I thought that was referring to the Münster Curry Compiler.
 
Ah
I don't actually know.
 
@Potato44 No, that wad about PAKCS. Unfortunately, neither of the three compilers I tried worked.
 

« first day (393 days earlier)      last day (2101 days later) »