Please wait...
That's the installer by NullSoft, right?
I had to build an installer for our client tool using that, because fucking MSI doesn't let you create decent installation packages that can truly handle 64-bit vs. 32-bit environments, where you build one installer that installs appropriately for either operating system.
So, I settled on NullSoft's thing. It can. I sorta hate it, because it requires you learn a scripting language they invented, which to me is a step backwards, but there you go.
I wouldn't call myself a guru with it, but maybe I can help you.
ah, yes thats great. Yes, its that Nullsoft thing.
i've got this template: https://github.com/arangodb/arangodb/blob/devel/Installation/Windows/Templates/NSIS.template.in
(https://docs.arangodb.com/cookbook/CompilingUnderWindows.html is the transscript howto generate it all)
and it installs to the systems startmenu like this:
/cygdrive/c/Users/All Users/Microsoft/Windows/Start Menu/Programs/ArangoDB 2.5.0-beta1
I've learned that one should call
SetShellVarContext all
so they get removed on uninstall again... I'm trying that, but the directory is not removed :-(
the *F*n compile process takes half an hour, and I can't seem to circumvent it, so the experience is pretty frustrating...
Uh, yeah, my installs are far simpler, so it doesn't take me as long to compile their stupid script.
Wow, yes, that's a serious script. I'll see if there's anything glaring concerning SetShellVarContext all .
Heh...
Push "this is a long ass string"
Push "ass"
Man, I love reading some of the colorful comments one can find in this sort of stuff.
Tue Mar 10 2015 10:15:23 EDT from fleeb @ Uncensored
Wow, yes, that's a serious script. I'll see if there's anything glaring concerning SetShellVarContext all .
Well, its that cmake has to recompile arangodb in order to generate the package...
I wasn't able to bypass that...
It's hard for me to tell, but I wonder if this is the problem:
http://nsis.sourceforge.net/Docs/Chapter4.html#setshellvarcontext
Note that it says 'if used in uninstaller code, this will only affect the uninstaller. To affect both, it needs to be used in both.'
So it's possible that you need to call this function during the uninstall portion of the script according to how it was used on the install portion of the script.
hm, it seems it doesn't reach the SetShellVarContext - printf debugging is awsome with 20 minutes for each shot - NOT. :-(
Fucking hell... if it weren't for the fact that the damned thing can write for both 64 and 32 bit machines, I would pitch the damned thing.
Doesn't reach it... I wonder if it's outside a function or something.
I do recall finding the process generally tedious. But, are you unable to perhaps create a lighter version of the install (with simple 1k files) to test setup a little faster?
*GNA NSIS* another Constant looking thing that flips depending on the shellcontetxt...
http://nsis.sourceforge.net/Reference/WriteRegExpandStr
If root_key is SHCTX or SHELL_CONTEXT, it will be replaced with HKLM if SetShellVarContext is set to
all and with HKCU if SetShellVarContext is set to current.
Deletes my Startmenu now on uninstall.
lol MSVC.You get this if there is another compile running (i.e. by the jenkins user) on the same server.
LINK : fatal error LNK1101: incorrect MSPDB120.DLL version; recheck installation of this product [C:\b\workspace\Contineous_W32_build\Build64\arangosh\arangosh.vcxproj]
Windows is peculiar about those All User vs. the installing user type crap.
It's just more of the fucking shell games Microsoft plays with regards to securing the operating system from hacking. The more byzantine Microsoft gets, the shittier the engineer's experience with Windows.
well, after some reading and lots of fiddling...
got it working the way I want it to.
The worse thing about NSIS probably realy is, that its got its completely own language not similar to anything, so once two weeks are passed, $brain drops everything and once another problem arises you start out again at zero...
Yeah, that's totally it.
If they'd only use something like lua or javascript, it wouldn't be so bad, but... ugh...
At long last!
[ http://www.theregister.co.uk/2015/04/15/ncr_android_atm ]
NCR is ditching Windoze on its ATM's, and replacing it with Linux. Interestingly, it's going to be an Android build, which seems to make sense, since Android is fast becoming the universal lightweight client-side operating system.
(Commie Comrade Richard Stallman would prefer if you called it "GNU/Android")
This is a great move. I wish them much success and that the rest of the industry will follow.
ATM's (and banks) are finally catching up to where Autozone was back in the early 2000's
http://en.wikipedia.org/wiki/SCO%E2%80%93Linux_controversies#SCO_v._AutoZone
Can't find what I am looking for with the court rulings, but the Autozone kicked some ass back in the day apart from that short write-up that makes it just a blocking move. P.J. is my only hope! I know I read some Slashdot back in the day to confirm it.
Sad that I can't find it now with a quick search. It is almost like some force is out there... controlling my searches and limiting them somehow, but I know, that does not make scene, so I continue about my business.
I just hope the ATM / bank folks don't pay the extortion of the M$ Android fees for "royalties" and other stepping on toes tax.
Yup. There's really no difference. Microsoft funded the SCO controversy; the only difference this time around is that they're attacking Linux out in the open instead of using SCO as a proxy.