« first day (2573 days earlier)      last day (2280 days later) » 

12:11 AM
@sp_BlitzErik I'll answer you to the extent I am bound by NDA ;)
 
What exactly does that say?
You should probably just post that on a pastebin so we know what not to ask.
It'll save Microsoft billions and stuff.
Like, how would we know whether or not you can publicize the trace flags?
Did they tell you that one explicitly, or are you just assuming they want that protected by a secret handshake?
I mean, if I was Microsoft I would want EVERYONE to know everything, so they must have given you some guidance about what is secret-handshake and what is declassified for public use?
 
@EvanCarroll The general answer is you can't.
@EvanCarroll Haha, basically if you don't see it in blogs or forums then it isn't public knowledge.
 
@SeanGallardy-Microsoft I think it would genuinely help us as outsiders if we had a reasonable line to stop asking
 
The gist would be
1) I can't share code
 
Or like, what kind of alcohol/clear heel would blur that line
 
12:19 AM
2) I can't tell you calls, defined values, algorithms
 
You mean you can't tell us the value, or what it does?
 
3) I can't tell you implementation details
So like, for example...
What does xe_This_isnt_ducmented do?
spelling mistakes and all
I'll go look it up, and as long as it isn't something horrible I'll give you the broad picture of what it does
I won't tell you where it fires, how it fires, or internally what it hits
Another example...
 
But I mean you can't even give the broad picture of the trace flags right?
 
There are a lot of trace flgs
 
Presumably you'd have to go knock on the door of the Microsoft Information Security Czar, or MS Genius or whatever they're call them.
 
12:22 AM
Example 2: Is there a trace flag to let me see the output of X?
If that TF isn't documented or otherwise mentioned, unless you're at the point of opening up a case with support where they would tell you to put it in (which I can check internally) then I can't say anything
I might hint that one exists jus to see you wiggle
I can paint broad strokes
I can say 0 on anything security related
 
Meh security
 
I know, Erik hates security
think about your social security numner - would you tell it to me?
Probably not.
 
You can't even tell me how to turn on Bit Locker, why would I trust you with anything important?
 
bwahahhahahah
 
12:24 AM
it works for me
 
You can have my social security number, hope you like debt
 
must be a PEBCAK error
I'll add more to it for sure
 
you want security, use Windows ME
 
if we're listing Os's that are 19 years old...
I'm sad that kids today won't get to discover the joys of windows 95
 
Yo
BeOS was the shit
That's all
 
12:27 AM
Here's an example of what I can do: dba.stackexchange.com/questions/154168/…
 
PLEASE UPVOTE ME MY NAME IS SEAN
 
lol
giving examples
here's another example, I can answer but I can't give you implementation: dba.stackexchange.com/questions/162140/…
 
not falling it for it.
 
lol someone did upvote me
 
Let's ask this the sweep that reveals the Trace Codes, why do you feel that is insufficient to find all of the trace codes?
And would a better sweep be doable
 
12:31 AM
Man sysprocesses has so much good stuff in it
 
@sp_BlitzErik I CHALLENGE YOU TO A GENTLEMANS DUEL
I SAID GOODDAY
 
Also I'm surprised you didn't link to my question about DMV values resetting
 
oh yeah that's a good one too
let me find it..
8
A: Do values in `sys.dm_os_wait_stats` reset to 0 or stop accumulating when they hit the max value?

Sean Gallardy - MicrosoftAccording to Docs: In SQL Server the wait-time counters are bigint values and therefore are not as prone to counter rollover as the equivalent counters in earlier versions of SQL Server. I've seen this with earlier version of SQL Server where rollover happened and you'd get negative num...

That's a good example because it's documented in BOL
So I can tell you stuff because it's already documented
I can tell it in slightly more detail that normal
@EvanCarroll example?
 
@SeanGallardy-Microsoft most of that was just straight up wrong. ;)
USELESS distractions
 
LMFAO
Think of a TF like this...
you enable it which just sets a value in memory
that value isn't checked
 
12:37 AM
How the fuck would that be useful?
 
it's an obfuscation
so that people can't just enabled TF's and see if it works or not
 
What is this 12 dimensional soduku
 
That doesn't make any sense.
 
eventually if it needs to branch from a tf...
 
12:38 AM
then it has to check to branch?
 
we'll see if an internal item is set
 
right, exactly
 
if it is we'll do something TF related, otherwise it'll continue on
 
right exactly
so it is checked.
 
it's not checked when you enable it
it's only checked when the code path is hit
 
12:39 AM
exactly
 
these are all things @PaulWhite already said
 
But he's not right for the same reason..
it's compiled code.
what makes you think that the code path has to be hit to determine what it's checking against.
 
I mean is my model of how compilations work off or something
an executable has all code paths, hit and not hit.
 
if it's compiled into the code
 
12:41 AM
that's what I'm saying, it has to be.
otherwise we're back at impossible magic
 
think about that, long and hard...
 
you have your own compiler inside SQL Server that dynamically inserts TF and you have a hooking mechanism that checks for dynamically compiled code?
 
We do have a compiler inside of SQL Server (Hekaton) but that's not it
actually, compiler and linker
not all builds of SQL Server have all TF's
 
yes, I know. but that's for something different.
the dll generating garbage on server start
 
12:43 AM
CUTE.
sure someone thought that was a fucking brilliant idea.
 
Assume you compiled all your flags in, right?
There is nothing stopping you from doing something like...
 
We have a query-compiler, and we need to market another compiler. so let's compile the queries before they're run rather than compiling them once when we get the query and caching them. Yeah! That's a great idea guyz.
 
IF(FLAG_1 is on){IF(Flag_2 is on){Check flag3}}
so if you turn flag 3 on but never flag 1 or flag 2...
you'll never hit the code
 
I'm soo confused here, You don't have to "hit" the code why are you paul hung up on that?
 
You do... though...
that's the onlytime it's checked
 
12:46 AM
Yes, if by checked you mean executed.
 
Who cares about execution I'm not stepping through with a debugger, I'm examining the executable
 
@sp_BlitzErik remember that udf stats thing?
@EvanCarroll examine away, my friend
 
if you disable constant folding, b=0; if(false){if(b==1){die)}, you'll still see the b==1 in the compiled executable
 
... and you have to know to look for b.
 
12:47 AM
that doesn't go away because the code path isn't hit. (assuming you disable constant folding)
Right well that the question how do I know what to look for.
For that all I have to do is watch what is being set when I call a single traceflag
 
if you say so
 
@SeanGallardy-Microsoft of course
 
I say so!
 
@sp_BlitzErik So, I got a response...
I am currently on vacation and will look into this as soon as I can.
haha
 
I hope I'm not wrong, but if I am I don't get where I am wrong.
 
12:49 AM
but she agrees something is off
 
and I've compiled tests that demonstrated this.
 
I think there would generally be less fetishizing of trace flags and weird insider tricks if more things were plainly exposed to end users without needing them or debuggers or thousands of insane DMV queries
 
I just need to get it working on something that is not ideal, like SQL Server.
 
@EvanCarroll You're confusing compilation with checking
@sp_BlitzErik Agreed.
 
@sp_BlitzErik totally agree. I would otherwise not care at all but it's fun like a puzzle.
 
12:50 AM
Especially with basic stuff. You have extended events out the wazoo but who even knows when to use what
 
TBH most are just for CSS
so we can debug without attaching debuggers
Most TF's aren't even helpful
 
Agreed, but it would be nice if everyone knew that
 
@EvanCarroll I like puzzles!
 
I mean think about how the undocumented dbcc commands used to seem like a big deal
 
Quit Microsoft and join me!
 
12:55 AM
Actually I love board games
I like Texas!
What's the offer?
;)
You already promised to take me from 0 to hero
fo' free
 
all of the trace flags! FREE.
if we win.
 
if we don't we win we raid Microsoft and send the filing cabinets to wikileaks.
we know the tracecodes are in filing cabinets
 
Are you stroking out again?
 
1:33 AM
It appears we've lost him
 
 
2 hours later…
3:44 AM
NEVER
 
 
4 hours later…
8:06 AM
Morning 📖
 
 
3 hours later…
10:57 AM
m
 
 
5 hours later…
4:17 PM
It's nice when after writing the whole answer you notice it is about oracle
 
5:14 PM
Anybody alive?
How should I execute my_pivot function?
 
HALLO MY PEOPLE
MY @McNets!!!
 
@EvanCarroll Hello, are you good with Oracle?
 
5:38 PM
noap
 
o
k
 
 
2 hours later…
8:24 PM
Let's learn radare on SQL Server!
@SeanGallardy-Microsoft better practice your secret handshakes before I get BETTER at them. ;)
 
8:59 PM
not a complete answer to your question however…
 
@JackDouglas Thank you"
ok let me see
 
more helpful would be if dbfiddle would display DBMS_OUTPUT output
one for the todo list
 
@JackDouglas I see, I can't get a rowset as a result like in SQL Server.
I won't post an answer. Just trying to get it working.
 
@JackDouglas can I use a temp table or similar?
 
9:05 PM
which question is it?
 
no there isn't a way to do that
 
A dinamyc pivot, however I'm not sure about the desired result.
 
it's ok if you stay inside the pl/sql context — that's basically what a weak ref cursor is
untyped I mean
do you know about INSERT ALL? dbfiddle.uk/…
 
no
ok, thank you
 
9:13 PM
0
Q: Oracle function for dynamic sql pivoting

RidRoidI found this amazing function in this article. the function turns a set of values in one of the columns into the column headings : CREATE OR REPLACE type PivotImpl as object ( ret_type anytype, -- The return type of the table function stmt varchar2(32767), fmt varchar2(32767), cu...

 
HALLO JACK. Merry Saturday.
 
I'm reading, but I better stay on MSSQL.
 
yes it's a bit mental
Hi @Evan
 
10:13 PM
@EvanCarroll No Olly/edb or clang++?
I had higher hopes for you...
 
10:23 PM
@SeanGallardy-Microsoft this isn't my thing man. I'm only getting into this because those methods suck, and I like learning.
This is wayyyy outside of my expertise, currently.
But a couple of months. We'll see!
However, as far outside of my expertise I know it's possible and that you and @PaulWhite are wrong. ;)
Paul should 500pt bounty that question. =)
I've done lots of static analysis on file formats, and the like, and this has to be doable. I think it should be <10 lines in this tool
This is really fucking cool tool.
 
 
1 hour later…

« first day (2573 days earlier)      last day (2280 days later) »