[Opensim-dev] My new LSL compiler project
Eric
e800 at gmail.com
Fri Nov 13 23:17:23 UTC 2015
Just an update on this.
I recently corrected a pretty large issue with modifying operations on
vector/rotation components being erroneously optimized away in certain
cases.
The expression node property that recursively checks the an expression tree
for expressions with side effects was failing to recognize that result of
the '.' operator access node could be modified by certain assignment and
postfix/prefix operations.
Therefore statements like:
vec.x *= 2; or vec.y++; (ect..)
Were interpreted as having no side effects.
I found this problem from the MONO tests Blake pointed out on the wiki.
==
I've since put up a new release, the editor now contains a working config
pane for the compiler in the settings window, and uses a LibLSLCC binary
with the previously mentioned optimization fix.
The compiler options pane is only partially field validated (need to write
validation stuff for method signatures, class inheritance list, and
constructor signature settings still)
The OpenSim fork has been updated as well to contain the fix.
I have also put some work into the experimental branch on my fork that
attempts to improves casts from strings to floats in all runtime types. It
adds formating support for NaN, Infinity (both pos and neg) as well as
preservation and printing of negative signed zeros. hex string to float
parsing has also been added.
The changes to the runtime types in the experimental branch have not been
heavily tested yet, but seem to fix most of the formating/cast from string
issues present in the compiler compliance test from the wiki.
Everything's still overflow checked, so that's still going to stop the
overflow tests from working of course, and quaternions still have
normalization differences (that's not really a correctable thing)
On Nov 9, 2015 5:20 PM, "Eric " <e800 at gmail.com> wrote:
> TIL that that ([0,0] != [0,0]) is (length(left) - length(right)).
> interesting.
>
> On Mon, Nov 9, 2015 at 3:47 PM, Melanie <melanie at t-data.com> wrote:
>
>> The [1] != [2,3] test would yield 1 because the c# notion of true is
>> 1, not -1.
>>
>> This can't be fixed because it's a Mono thing. A compiler cold work
>> around it by identifying expressions yielding a boolean value and
>> wrapping them, like so: ([1] != [2,3] != 0 ? -1 : 0)
>>
>> But that is ugly and inefficient.
>>
>> - Melanie
>>
>> On 09/11/2015 22:36, Eric wrote:
>> > Blake,
>> >
>> > I found the LSL language test on the wiki when I first started working
>> on
>> > the library,
>> > I forgot where I got them and went looking for the other ones but could
>> not
>> > find them again.
>> >
>> > These in particular:
>> >
>> > http://wiki.secondlife.com/wiki/LSL_Language_Test
>> > http://wiki.secondlife.com/wiki/LSL_Language_Test_2
>> >
>> > There are a few current failures in the first test:
>> >
>> > The first is due to quaternion division producing different (but valid)
>> > results in OpenSim on line 436
>> > The second is on line 471-473 and it is caused by an overflow exception
>> > being thrown
>> > The third is on line 480-482 and it is also caused by an overflow
>> exception.
>> >
>> > The rest of the failures are caused by the Float equality delta tests,
>> and
>> > comparing vectors/rotations cast to strings. (their formatting is
>> slightly
>> > different in opensim)
>> >
>> >
>> > There are several in the second, due to vectors and rotations in
>> OpenSims
>> > runtime converting into strings with spaces after the commas.
>> > In secondlife there are no spaces in the list of components that make
>> up a
>> > vector or rotation when its ToString()'d
>> >
>> > And an instance of: [1] != [2,3] (1 expected -1), on line 224 of
>> > LSL_Language_Test_2 which is a runtime related thing I am curious to
>> > investigate.
>> >
>> >
>> > I am unsure if I should improve the actual runtime types, or fix
>> > differences generating workaround code.
>> >
>> > I am inclined to believe that the quaternion runtime type should
>> probably
>> > not be messed with besides fixing the ToString() formatting,
>> > I am not sure what opinions there are on modifying the other existing
>> types.
>> >
>> > ==
>> >
>> > All benchmark scripts on the page work and produce identical output to
>> > second life.
>> >
>> >
>> > LSL_Library_Call_Test_1 failed due to OpenSim using void for the return
>> > type of llGiveMoney();, and also due to erroneous function parameters in
>> > the script causing exceptions.
>> >
>> > LSL_Library_Call_Test_2 compiles and fails due to erroneous parameters
>> to
>> > LSL functions causing exceptions, which halt the script.
>> >
>> > Event_test_script fails due an issue with the order in which events are
>> > called by OpenSim (I'm not exactly sure why).
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> > On Mon, Nov 9, 2015 at 12:39 PM, Michael Emory Cerquoni <
>> > nebadon2025 at gmail.com> wrote:
>> >
>> > > doesnt LL use their own forked version of mono?
>> > > On Nov 9, 2015 10:38 AM, "Blake" <techplex.engineer at gmail.com> wrote:
>> > >
>> > >> Eric,
>> > >>
>> > >> Have you seen the Mono test scripts that SL used when they change
>> > >> compilers from the legacy LSL to the Mono based one?
>> > >> They might be a good benchmark. I know that a bunch of the test fail
>> in
>> > >> OS.
>> > >>
>> > >> http://wiki.secondlife.com/wiki/Mono#Testing
>> > >>
>> > >> Best,
>> > >> Blake
>> > >>
>> > >> On Mon, Nov 9, 2015 at 1:29 PM, Eric <
>> e800 at gmail.com>
>> > >> wrote:
>> > >>
>> > >>> If yall believe that reverse compatibility should be thrown out the
>> > >>> door, it makes everything much simpler.
>> > >>>
>> > >>> I was under a preconceived notion for a bit that any sort of script
>> > >>> breakage might cause the compiler to be a no go.
>> > >>>
>> > >>> Fortunately all changes I mentioned are easy to make and reverse
>> due to
>> > >>> the, way the expression type evaluation and checking logic is
>> segregated
>> > >>> out in the library.
>> > >>>
>> > >>>
>> > >>> At the moment I am tidying up some things and updating the build
>> system,
>> > >>> mainly to remove the platform specific binaries from the installer
>> and
>> > >>> binary releases;
>> > >>> Also I am working on a settings UI for the editor.
>> > >>>
>> > >>> I've just removed the tiny change from the library that allows for
>> the
>> > >>> compatibility options previously mentioned.
>> > >>>
>> > >>>
>> > >>>
>> > >>>
>> > >>> I have experienced a few problems as you mentioned with the default
>> > >>> XEngine compiler, I have put at least one script through it that
>> caused a
>> > >>> stack heap
>> > >>> and server crash in its current state. It is this one:
>> > >>> http://wiki.secondlife.com/wiki/BigNum, however I have not
>> > >>> investigated why it happens yet. It does not
>> > >>> happen when LibLSLCC is used for code generation.
>> > >>>
>> > >>> I am pretty sure the crash happens in the default XEngine compiler
>> > >>> itself and not in the running code.
>> > >>>
>> > >>> Examples of either type of problem (in running code, or especially
>> > >>> compiler related) would be helpful in testing LibLSLCC.
>> > >>>
>> > >>>
>> > >>>
>> > >>>
>> > >>>
>> > >>>
>> > >>>
>> > >>> On Mon, Nov 9, 2015 at 10:25 AM, Zadark Portal <
>> zadarkportal at gmail.com>
>> > >>> wrote:
>> > >>>
>> > >>>> Hello Eric
>> > >>>>
>> > >>>> I agree with the previous comment.
>> > >>>>
>> > >>>> This contribution is such that we are now considering reintroducing
>> > >>>> Opensim as an entry platform to scripting. Previously we found it
>> > >>>> impossible to support larger class sizes due to the empirical
>> nature of
>> > >>>> novices producing non deterministic behaviours.
>> > >>>>
>> > >>>> If when considering further developments may I suggest trapping
>> > >>>> potentially dangerous (Opensim crash) scripted constructs. Maybe
>> this is a
>> > >>>> tall ask, but enormously beneficial to simulator operators managing
>> > >>>> experimental behaviour of users attracted to Opensim. I have not
>> included
>> > >>>> examples but will to do so at your request.
>> > >>>>
>> > >>>> It will be a couple of weeks before I can introduce the
>> functionality
>> > >>>> to a class and comment further, but whatever the future for this
>> > >>>> contribution with regards to core we will continue make full use.
>> > >>>>
>> > >>>> Regards
>> > >>>> Z
>> > >>>>
>> > >>>> On 9 November 2015 at 14:20, Melanie <melanie at t-data.com> wrote:
>> > >>>>
>> > >>>>> Please don't do all these things.
>> > >>>>>
>> > >>>>> OpenSim is alpha software and there is no need to make the mistake
>> > >>>>> other projects made and perpetuate bugs in the name of
>> compatibility.
>> > >>>>>
>> > >>>>> The cast to list is there because, afaik, in SL the operation
>> list +
>> > >>>>> scalar is legal and means to append an element. It's not meant to
>> > >>>>> allow the other side effects you listed and they should be
>> illegal.
>> > >>>>>
>> > >>>>> There isn't a way to do these tests on crossings. That's also not
>> > >>>>> necessary as the better compiler will be used by all in short
>> order.
>> > >>>>> People will fix their scripts and everyone will be happy again.
>> > >>>>>
>> > >>>>> Let's not breed bugs, but let's squish them instead.
>> > >>>>>
>> > >>>>> - Melanie
>> > >>>>>
>> > >>>>> On 09/11/2015 11:50, Eric wrote:
>> > >>>>> > You're right actually.
>> > >>>>> >
>> > >>>>> > Since stuff like old LSO list/string optimization hacks were not
>> > >>>>> possible
>> > >>>>> > beforehand while using XEngines compiler,
>> > >>>>> > there is probably not to many people with code that's going to
>> behave
>> > >>>>> > incorrectly with the evaluation order of binary
>> > >>>>> > operations switched. The scripts that happen to break are doing
>> > >>>>> something
>> > >>>>> > pretty wacky/unnecessary.
>> > >>>>> >
>> > >>>>> > There's not any micro optimizations that I know of relying on
>> the
>> > >>>>> old left
>> > >>>>> > to right evaluation order either.
>> > >>>>> >
>> > >>>>> > ==
>> > >>>>> >
>> > >>>>> > That being said, there are many things the XEngine default
>> compiler
>> > >>>>> will
>> > >>>>> > allow to compile that are infact
>> > >>>>> > syntactically incorrect in LSL. stuff like negating
>> > >>>>> vectors/rotations in
>> > >>>>> > global variable declarations
>> > >>>>> > as well as doing cast in static context. In general the default
>> > >>>>> XEngine
>> > >>>>> > compiler's rules for global variable
>> > >>>>> > declaration expression content are very lax in comparison to the
>> > >>>>> secondlife
>> > >>>>> > compiler, and that is just one example.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > LibLSLCC's goal to mimic the secondlife compiler's rules
>> exactly, in
>> > >>>>> its
>> > >>>>> > default configuration.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > From my understanding a script moving to new a region is going
>> to be
>> > >>>>> > recompiled if the destination sim does not allow the script
>> binary
>> > >>>>> > to transfer over and be executed. This will be true in most
>> cases
>> > >>>>> because
>> > >>>>> > the default settings disallow it, and it's unsafe to enable.
>> > >>>>> >
>> > >>>>> > This would cause a problem for scripts from another region that
>> > >>>>> LibLSLCC
>> > >>>>> > deems to contain syntax errors.
>> > >>>>> >
>> > >>>>> > I may to have to dig into OpenSim to see how region crossing for
>> > >>>>> scripts
>> > >>>>> > work; to see if I have to make the re-compile sensitive
>> > >>>>> > to which compiler was used on the other sim, regardless of what
>> > >>>>> version of
>> > >>>>> > OpenSim it is running.
>> > >>>>> >
>> > >>>>> > I imagine this would be a requirement to connect a sim to a grid
>> > >>>>> containing
>> > >>>>> > regions not using my compiler addon.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > ==
>> > >>>>> >
>> > >>>>> > I tested the library build on OS X Snow leopard 10.6 and OS X El
>> > >>>>> Capitan
>> > >>>>> > 10.11.
>> > >>>>> >
>> > >>>>> > Building LibLSLCC-NoEditor.sln works as intended.
>> > >>>>> >
>> > >>>>> > The command line program lslcc that's part of LibLSLCC works
>> fine on
>> > >>>>> both
>> > >>>>> > and has no problem generating code.
>> > >>>>> >
>> > >>>>> > LibraryDataScrapingTools works on OSX 10.11 even with its
>> WinForms
>> > >>>>> dialogs,
>> > >>>>> > and the OpenSim fork works fine to.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > However, when I tested with OSX 10.6 I ran into an issue where
>> mono
>> > >>>>> > complained about missing System.Drawing.GDIPlus.
>> > >>>>> >
>> > >>>>> > I'm sure there's probably a way to fix this but I did not look
>> into
>> > >>>>> it to
>> > >>>>> > deeply as I was using a borrowed computer for a few minutes,
>> > >>>>> > that error prevented both LibraryDataScrapingTools and OpenSim
>> from
>> > >>>>> > starting.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > ==
>> > >>>>> >
>> > >>>>> > As of now I have added a setting to allow implicit cast of all
>> types
>> > >>>>> into
>> > >>>>> > list within a function
>> > >>>>> > calls parameter expressions, as Zadark had mentioned.
>> > >>>>> >
>> > >>>>> > As it turns out, all of OpenSim's C# runtime types contain an
>> > >>>>> implicit
>> > >>>>> > conversion operator to list.
>> > >>>>> >
>> > >>>>> > It happens regardless of context.
>> > >>>>> >
>> > >>>>> > So The default XEngine compiler allows stuff like this too:
>> > >>>>> >
>> > >>>>> > list a = "";
>> > >>>>> >
>> > >>>>> > list b = <0,0,0>;
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > I will probably change the setting to just allow implicit list
>> cast
>> > >>>>> > everywhere as an option,
>> > >>>>> > and maybe add some syntax related compatibility options for
>> global
>> > >>>>> > variables.
>> > >>>>> >
>> > >>>>> > In the long run, if I can figure out how to make the compiler
>> addon
>> > >>>>> aware
>> > >>>>> > of re-compiles due to region
>> > >>>>> > crossing, some syntax in-compatibility will not be a major
>> issue for
>> > >>>>> > scripts that you cannot modify.
>> > >>>>> >
>> > >>>>> > If I could somehow select to re-compile them with the default
>> XEngine
>> > >>>>> > compiler (if that's what they were originally compiled with),
>> > >>>>> > then they will not be subject to LibLSLCC's stricter syntax
>> checking.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > ==
>> > >>>>> >
>> > >>>>> > I have been working on a settings panel for the editor so I can
>> > >>>>> easily test
>> > >>>>> > different
>> > >>>>> > compiler configurations, and segregating some of the editor
>> code out
>> > >>>>> into
>> > >>>>> > different assemblies;
>> > >>>>> > it will take me a little bit to finish.
>> > >>>>> >
>> > >>>>> > I'm currently working out of the development branch on github,
>> I'll
>> > >>>>> put up
>> > >>>>> > a new editor installer
>> > >>>>> > when I get the configuration stuff all worked out.
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > On Sat, Nov 7, 2015 at 8:46 AM, Melanie <melanie at t-data.com>
>> wrote:
>> > >>>>> >
>> > >>>>> >> I would be against making the evaluation order optional. It
>> has long
>> > >>>>> >> been recognized as a bug in XEngine's compiler and people
>> porting
>> > >>>>> >> scripts from SL have been advised to remove evaluation order
>> > >>>>> >> dependent code but not write any code that depends on the wrong
>> > >>>>> >> XEngine eval order. Therefore, if this compiler breaks content
>> that
>> > >>>>> >> content needs to be fixed, not a setting changed for
>> convenience to
>> > >>>>> >> perpetuate the bad behavior.
>> > >>>>> >>
>> > >>>>> >> - Melanie
>> > >>>>> >>
>> > >>>>> >> On 07/11/2015 15:36, Eric wrote:
>> > >>>>> >> > Thanks Again Zadark
>> > >>>>> >> >
>> > >>>>> >> > This should be easy to fix, I need to change the type
>> validation
>> > >>>>> system
>> > >>>>> >> > just little so the syntax tree can be built with these types
>> of
>> > >>>>> implicit
>> > >>>>> >> > conversion present.
>> > >>>>> >> >
>> > >>>>> >> > The code generator can then be made to generate the cast when
>> > >>>>> necessary,
>> > >>>>> >> > since it will pass syntax checking.
>> > >>>>> >> >
>> > >>>>> >> > I think I am going to implement it as a compatibility option
>> in
>> > >>>>> the
>> > >>>>> >> > compiler that you can turn on or off, depending on if you
>> want
>> > >>>>> the linden
>> > >>>>> >> > compilers more strict type checking behavior.
>> > >>>>> >> >
>> > >>>>> >> > I should also probably make order of evaluation correction on
>> > >>>>> binary
>> > >>>>> >> > operations something you can disable if needed, since I can
>> see
>> > >>>>> that
>> > >>>>> >> > causing compatibility problems somewhere too.
>> > >>>>> >> >
>> > >>>>> >> > Side Note: I have not tested on Mac yet, but I plan to do so
>> > >>>>> today if I
>> > >>>>> >> > can get time to, sorry.
>> > >>>>> >> > On Nov 7, 2015 7:42 AM, "Zadark Portal" <
>> zadarkportal at gmail.com>
>> > >>>>> wrote:
>> > >>>>> >> >
>> > >>>>> >> >> Hello Eric,
>> > >>>>> >> >>
>> > >>>>> >> >> Very useful, thankyou.
>> > >>>>> >> >>
>> > >>>>> >> >> Caveat
>> > >>>>> >> >> XEngine appears to cast library function parameters to the
>> > >>>>> relevant
>> > >>>>> >> type,
>> > >>>>> >> >> though I can see this handy for novice scriptwriters it does
>> > >>>>> encourage
>> > >>>>> >> some
>> > >>>>> >> >> less than elegant techniques
>> > >>>>> >> >>
>> > >>>>> >> >> Some popular scripts freely available to OpenSim use this
>> > >>>>> feature:
>> > >>>>> >> >> For instance:
>> > >>>>> >> >> llListFindList takes 2 parameters of the type List, when in
>> fact
>> > >>>>> the
>> > >>>>> >> >> existing compiler permits any type (as far as we have
>> tested).
>> > >>>>> >> >>
>> > >>>>> >> >> example... integer bah = llListFindList( blah1,
>> > >>>>> llList2String(blah2,
>> > >>>>> >> >> blah3)); // compiles and kind of works.
>> > >>>>> >> >>
>> > >>>>> >> >> This of course fails when compiled with LSLcc, the
>> explanation
>> > >>>>> is clear
>> > >>>>> >> >> and precise so easy to fix (enable compile) with a local
>> cast.
>> > >>>>> >> >>
>> > >>>>> >> >> Z
>> > >>>>> >> >>
>> > >>>>> >> >>
>> > >>>>> >> >> On 5 November 2015 at 12:52, Jak Daniels <jak at ateb.co.uk>
>> wrote:
>> > >>>>> >> >>
>> > >>>>> >> >>> Hi Eric,
>> > >>>>> >> >>>
>> > >>>>> >> >>> yes I can confirm now that it does open and compile
>> correctly
>> > >>>>> in VS C#
>> > >>>>> >> >>> 2010 express now. Thank you.
>> > >>>>> >> >>>
>> > >>>>> >> >>> I'm now installing VS2015 in a win10 virtual machine so I
>> don't
>> > >>>>> clobber
>> > >>>>> >> >>> my SSD, so I can compile and play with the LSL editor.
>> > >>>>> >> >>>
>> > >>>>> >> >>> kind regards
>> > >>>>> >> >>> Jak
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>> On 04/11/2015 06:10, Eric wrote:
>> > >>>>> >> >>>
>> > >>>>> >> >>> Let me just make a slight correction to my last message,
>> and
>> > >>>>> add some
>> > >>>>> >> >>> info...
>> > >>>>> >> >>>
>> > >>>>> >> >>> After removing the "<LangVersion>5</LangVersion>" stuff
>> from the
>> > >>>>> >> >>> non-editor related .csproj files,
>> > >>>>> >> >>>
>> > >>>>> >> >>> I found that VS2010 is actually able to open the
>> > >>>>> >> >>> "MonoDevelop-LibLSLCC.sln" solution file in the source
>> tree and
>> > >>>>> build
>> > >>>>> >> it
>> > >>>>> >> >>> without any problems.
>> > >>>>> >> >>>
>> > >>>>> >> >>> I'm going to be pushing the changes I made to the .csproj
>> files
>> > >>>>> >> shortly,
>> > >>>>> >> >>> I will probably rename "MonoDevelop-LibLSLCC.sln" to
>> something
>> > >>>>> like
>> > >>>>> >> >>> "LibLSLCC-No-Editor.sln" and update the build instructions
>> in
>> > >>>>> the
>> > >>>>> >> >>> README.md to specify that solution as the solution to use
>> for
>> > >>>>> building
>> > >>>>> >> the
>> > >>>>> >> >>> library on Mono.
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>> I have not tested that the editor build is working with
>> > >>>>> anything other
>> > >>>>> >> >>> than vs2015, but I will test with the prior versions
>> vs2012+
>> > >>>>> to see
>> > >>>>> >> what
>> > >>>>> >> >>> all
>> > >>>>> >> >>> works, then put some info in the README.md about what
>> IDE's the
>> > >>>>> Editor
>> > >>>>> >> >>> portion of the project can be built in.
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>> On Tue, Nov 3, 2015 at 11:53 PM, Eric <
>> > >>>>> e800 at gmail.com
>> > >>>>> >> >
>> > >>>>> >> >>> wrote:
>> > >>>>> >> >>>
>> > >>>>> >> >>>> Thanks for letting me know about this problem Jak. (See
>> the
>> > >>>>> bottom
>> > >>>>> >> for
>> > >>>>> >> >>>> TLDR :P)
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I am using vs2015 and WiX 3.10.1, it does indeed seem to
>> be a
>> > >>>>> problem
>> > >>>>> >> >>>> with vs2010 C#.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I was unable to find an updated download for vs2010 C# to
>> test
>> > >>>>> with,
>> > >>>>> >> >>>> only an SP1 ISO I found on stackoverflow:
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >>
>> > >>>>>
>> http://stackoverflow.com/questions/6871865/url-for-visual-studio-2010-express-iso
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> http://go.microsoft.com/?linkid=9709969
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I tried installing vs2010, however this version will not
>> > >>>>> attempt to
>> > >>>>> >> open
>> > >>>>> >> >>>> my solution files at all as it says they are incompatible.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I can however, load the .csproj files for certain
>> projects in
>> > >>>>> my
>> > >>>>> >> build.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I cannot get it to update from SP1, I am unsure if
>> updates are
>> > >>>>> still
>> > >>>>> >> >>>> available to download for vs2010.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> When I try to update, it just directs me to use windows
>> update
>> > >>>>> which
>> > >>>>> >> >>>> does not seem to find any updates related to it.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> === Why it's not loading ===
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> With vs2010 SP1, I was only able to load:
>> LibLSLCC.csproj,
>> > >>>>> >> >>>> lslcc_cmd.csproj and DemoArea.csproj.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> LibraryDataScrapingTools was incompatible because it was
>> > >>>>> building
>> > >>>>> >> >>>> against .NET 4.5.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I downgraded LibraryDataScrapingTools to .NET 4.0 and
>> > >>>>> re-installed
>> > >>>>> >> the
>> > >>>>> >> >>>> 'System.Data.Sqlite.Core' package from nuget using vs2015
>> in
>> > >>>>> order to
>> > >>>>> >> make
>> > >>>>> >> >>>> it load in vs2010.
>> > >>>>> >> >>>> I then made some code changes to make it compatible with
>> .NET
>> > >>>>> 4.0.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I will be committing the changes to
>> LibraryDataScrapingTools
>> > >>>>> tonight,
>> > >>>>> >> as
>> > >>>>> >> >>>> this project is supposed to be .NET 4.0 compatible.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> The LSLCCEditor related projects I am pretty sure are
>> > >>>>> incompatible at
>> > >>>>> >> >>>> the moment due to the .NET framework level they require as
>> > >>>>> well.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> The code base for LSLCCEditor requires a minimum of .NET
>> 4.5 to
>> > >>>>> >> compile,
>> > >>>>> >> >>>> and the code changes required to make the editor projects
>> > >>>>> >> >>>> compatible with .NET 4.0 are not as trivial as the
>> changes I
>> > >>>>> made to
>> > >>>>> >> the
>> > >>>>> >> >>>> LibraryDataScrapingTools project.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I am not sure if I want to limit the LSLCCEditor part of
>> the
>> > >>>>> project
>> > >>>>> >> to
>> > >>>>> >> >>>> a .NET 4.0 at the moment, as it is a Windows
>> > >>>>> >> >>>> only WPF application, and that sort of compatibility
>> profile
>> > >>>>> is not
>> > >>>>> >> >>>> necessary.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I also noticed when installing WiX v3.10.1 on a Virtual
>> > >>>>> Machine with
>> > >>>>> >> >>>> vs2010 installed that It does not try to register itself
>> with
>> > >>>>> vs2010
>> > >>>>> >> as a
>> > >>>>> >> >>>> project type.
>> > >>>>> >> >>>> I do not think the latest version of WiX installer
>> framework
>> > >>>>> supports
>> > >>>>> >> >>>> vs2010 anymore.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> === How I got LibLSLCC, LibraryDataScrapingTools,
>> lslcc_cmd and
>> > >>>>> >> DemoArea
>> > >>>>> >> >>>> to build. ===
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> After the code changes to LibraryDataScrapingTools and
>> > >>>>> downgrading the
>> > >>>>> >> >>>> .NET Framework level...
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I created a new vs2010 project and added (LibLSLCC,
>> > >>>>> >> >>>> LibraryDataScrapingTools, lslcc_cmd and DemoArea) as
>> existing
>> > >>>>> >> projects.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> When trying to build these projects I got the message
>> "Error 1
>> > >>>>> Invalid
>> > >>>>> >> >>>> option '5' for /langversion; must be ISO-1, ISO-2, 3 or
>> > >>>>> Default"
>> > >>>>> >> >>>> for all of the projects except 'DemoArea'
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I resolved this by manually editing the .csproj file for
>> each
>> > >>>>> >> >>>> non-building project in a text editor, removing the
>> > >>>>> occurrences of:
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> <LangVersion>5</LangVersion>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> That particular MSBuild setting is not necessary for
>> vs2015 or
>> > >>>>> Mono to
>> > >>>>> >> >>>> build the library anyway, so I am going to remove it and
>> > >>>>> commit the
>> > >>>>> >> changes.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> After that, LibLSLCC, LibraryDataScrapingTools, lslcc_cmd
>> and
>> > >>>>> DemoArea
>> > >>>>> >> >>>> were all able to build under vs2010.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> ===
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> That all being said, I will be adding a vs2010 compatible
>> > >>>>> solution
>> > >>>>> >> file
>> > >>>>> >> >>>> to the repository that contains Projects:
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> * LibLSLCC,
>> > >>>>> >> >>>> * lslcc_cmd
>> > >>>>> >> >>>> * LibraryDataScrapingTools
>> > >>>>> >> >>>> * DemoArea
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> So that you can at-least build the library, command line
>> > >>>>> compiler,
>> > >>>>> >> >>>> library data scraper and demo project with vs2010.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> Unfortunately making LSLCCEditor build in vs2010 would
>> require
>> > >>>>> some
>> > >>>>> >> non
>> > >>>>> >> >>>> trivial code changes, a new method of triggering the WiX
>> > >>>>> installer
>> > >>>>> >> build,
>> > >>>>> >> >>>> and a downgrade from .NET 4.5 to .NET 4.0 for the editor
>> > >>>>> related
>> > >>>>> >> >>>> projects.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> I'm not sure if I want to make these changes to the Editor
>> > >>>>> portion of
>> > >>>>> >> >>>> the project just for the sake of being compatible with
>> older
>> > >>>>> versions
>> > >>>>> >> of
>> > >>>>> >> >>>> Visual Studio.
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>> On Tue, Nov 3, 2015 at 10:05 AM, Jak Daniels < <
>> jak at ateb.co.uk
>> > >>>>> >
>> > >>>>> >> >>>> jak at ateb.co.uk> wrote:
>> > >>>>> >> >>>>
>> > >>>>> >> >>>>> Hi Eric,
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> Thank you for posting up your work to this group. This
>> all
>> > >>>>> looks very
>> > >>>>> >> >>>>> promising, and I was intrigued to give it a try.
>> > >>>>> >> >>>>> I checked out the source from github and I tried
>> building in
>> > >>>>> VS C#
>> > >>>>> >> 2010
>> > >>>>> >> >>>>> express... but only 3 of the projects will load.
>> > >>>>> >> LibraryDataScrapingTools,
>> > >>>>> >> >>>>> LSLCCEditor, LSLCCEditor.CompletionUI all say
>> (incompatible)
>> > >>>>> and
>> > >>>>> >> >>>>> LSLCCEditorInstaller says (unavailable) even though I
>> have
>> > >>>>> WIX 3.10.1
>> > >>>>> >> >>>>> installed.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> Is this a VS 2010 problem do you think?
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> I'm a bit loathed to install VS2015 as the last time I
>> > >>>>> installed
>> > >>>>> >> VS2013
>> > >>>>> >> >>>>> it just dumped everything onto my SSD c: drive and
>> filled it
>> > >>>>> up,
>> > >>>>> >> with no
>> > >>>>> >> >>>>> option to put things on drive D: my large harddrive. It
>> also
>> > >>>>> left a
>> > >>>>> >> mess
>> > >>>>> >> >>>>> behind (more than 50% of itself including MSSQL server
>> stuff)
>> > >>>>> when I
>> > >>>>> >> tried
>> > >>>>> >> >>>>> to uninstall it. VS2015 might be better now in this
>> respect,
>> > >>>>> but
>> > >>>>> >> knowing MS
>> > >>>>> >> >>>>> products probably not! I also found VS2013 to be way, way
>> > >>>>> slower
>> > >>>>> >> loading up
>> > >>>>> >> >>>>> large projects like OpenSimulator than VS2010. So before
>> I
>> > >>>>> take the
>> > >>>>> >> >>>>> possibly irreversible path of installing VS2015, is there
>> > >>>>> something
>> > >>>>> >> I can
>> > >>>>> >> >>>>> do to make it load in VS2010 express?
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> Thanks
>> > >>>>> >> >>>>> Jak
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> On 02/11/2015 19:24, Eric wrote:
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> Hello all.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> I am a bit new to OpenSim development (well at-least
>> sharing
>> > >>>>> stuff I
>> > >>>>> >> >>>>> have made..) but for a quite a while now I have been
>> working
>> > >>>>> on
>> > >>>>> >> >>>>> a new (BSD Licensed) Compilation/Code Generation
>> framework
>> > >>>>> for LSL,
>> > >>>>> >> >>>>> tailored towards usage with OpenSim. I thought I should
>> > >>>>> share it
>> > >>>>> >> >>>>> at this point of its development.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> This is pretty much a "full" or "true" compiler front
>> end.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> It's built on-top of ANTLR4 and ANTLR4's CSharp target.
>> > >>>>> >> >>>>> ANTLR4 however has been completely abstracted and the
>> library
>> > >>>>> >> provides
>> > >>>>> >> >>>>> its own Rich LSL Syntax Tree for users to deal with.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> My library includes an OpenSim code generation target
>> which I
>> > >>>>> have
>> > >>>>> >> >>>>> integrated into my OpenSim fork on GitHub, it's
>> implemented
>> > >>>>> as an
>> > >>>>> >> >>>>> optional compiler that you can enable in your
>> OpenSim.ini.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The Code validation step my library performs when
>> building a
>> > >>>>> syntax
>> > >>>>> >> >>>>> tree implements full front end syntax checking, dead code
>> > >>>>> detection,
>> > >>>>> >> the
>> > >>>>> >> >>>>> works.
>> > >>>>> >> >>>>> It also emits extended warning information that is
>> standard
>> > >>>>> to most
>> > >>>>> >> >>>>> compilers now days.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The OpenSim code generator I have written and included
>> with
>> > >>>>> the
>> > >>>>> >> library
>> > >>>>> >> >>>>> drastically improves compatibility with scripts written
>> for
>> > >>>>> >> SecondLife.
>> > >>>>> >> >>>>> Order of evaluation in generated code is correct for LSL
>> > >>>>> (Right to
>> > >>>>> >> >>>>> Left) among many other things.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> As an example, all of the encryption scripts you can
>> find on
>> > >>>>> the LSL
>> > >>>>> >> >>>>> wiki will compile correctly and execute with correct
>> behavior
>> > >>>>> using
>> > >>>>> >> my
>> > >>>>> >> >>>>> compiler.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The README.md for the project goes into a bit more
>> detail on
>> > >>>>> what all
>> > >>>>> >> >>>>> the library can do.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> My library LibLSLCC is on GitHub here:
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> <https://github.com/EriHoss/LibLSLCC>
>> > >>>>> >> >>>>> https://github.com/EriHoss/LibLSLCC
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The project includes an LSL Editor (Windows Only, I used
>> > >>>>> AvalonEdit)
>> > >>>>> >> >>>>> with the project that features code completion and syntax
>> > >>>>> >> highlighting.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> It can be used to test CSharp code generation for
>> OpenSim by
>> > >>>>> >> compiling
>> > >>>>> >> >>>>> LSL into C# using LibLSLCC, which can then be uploaded
>> to an
>> > >>>>> >> >>>>> OpenSim server with C# scripting enabled.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The library itself is cross-platform, but the editor and
>> > >>>>> editor
>> > >>>>> >> >>>>> installer are not. There's a separate project file for
>> > >>>>> building the
>> > >>>>> >> >>>>> library on
>> > >>>>> >> >>>>> mono with monodevelop/xbuild.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> ===
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The OpenSim fork that integrates my compiler is here:
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> <https://github.com/EriHoss/OpenSim_With_LibLSLCC>
>> > >>>>> >> >>>>> https://github.com/EriHoss/OpenSim_With_LibLSLCC
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> It includes a few minor bug fixes to XEngine and Runtime
>> > >>>>> Script
>> > >>>>> >> >>>>> functions.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> Such as the 'IdleTimeout' setting not being honored
>> properly.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> And llParseString2List using culture specific
>> comparisons,
>> > >>>>> causing it
>> > >>>>> >> >>>>> to misbehave when comparing Unicode characters on Mono.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> I have added new attributes to OpenSim's script module
>> > >>>>> >> >>>>> constants/functions and also to ScriptBaseClass.
>> > >>>>> >> >>>>> This is so LibLSLCC can de-serialize the classes into
>> library
>> > >>>>> data
>> > >>>>> >> >>>>> that's consumable by its code validator and code
>> generator.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> I have also made some slight changes to
>> IScriptModuleComms and
>> > >>>>> >> >>>>> ScriptModuleCommsModule.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> The old compiler in my fork works as it did before, so
>> you
>> > >>>>> can switch
>> > >>>>> >> >>>>> back and forth from LibLSLCC to the old compiler
>> > >>>>> >> >>>>> if you want without any problems.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> There are more details on the changes I have made in the
>> > >>>>> README.md.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> ==
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> I started working on this sometime early last year when I
>> > >>>>> broke my
>> > >>>>> >> >>>>> wrist and was unable to do much for a while, and
>> > >>>>> >> >>>>> I have only recently moved the code from my Git server
>> up to
>> > >>>>> GitHub.
>> > >>>>> >> >>>>> I'm going to continue improving this in my spare time,
>> > >>>>> >> >>>>> right now I am doing rolling releases versioned by date
>> > >>>>> anywhere from
>> > >>>>> >> >>>>> once every few days to a few times a day. I am also
>> keeping
>> > >>>>> my
>> > >>>>> >> OpenSim
>> > >>>>> >> >>>>> fork synced with the latest OpenSim commits.
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> Hopefully this will be useful, any feedback is
>> appreciated :)
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> _______________________________________________
>> > >>>>> >> >>>>> Opensim-dev mailing listOpensim-dev
>> @opensimulator.orghttp://
>> > >>>>> >> opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>> _______________________________________________
>> > >>>>> >> >>>>> Opensim-dev mailing list
>> > >>>>> >> >>>>> Opensim-dev at opensimulator.org
>> > >>>>> >> >>>>>
>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>>
>> > >>>>> >> >>>>
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>> _______________________________________________
>> > >>>>> >> >>> Opensim-dev mailing listOpensim-dev
>> @opensimulator.orghttp://
>> > >>>>> >> opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>> _______________________________________________
>> > >>>>> >> >>> Opensim-dev mailing list
>> > >>>>> >> >>> Opensim-dev at opensimulator.org
>> > >>>>> >> >>>
>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >> >>>
>> > >>>>> >> >>>
>> > >>>>> >> >>
>> > >>>>> >> >> _______________________________________________
>> > >>>>> >> >> Opensim-dev mailing list
>> > >>>>> >> >> Opensim-dev at opensimulator.org
>> > >>>>> >> >>
>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >> >>
>> > >>>>> >> >>
>> > >>>>> >> >
>> > >>>>> >> >
>> > >>>>> >> >
>> > >>>>> >> > _______________________________________________
>> > >>>>> >> > Opensim-dev mailing list
>> > >>>>> >> > Opensim-dev at opensimulator.org
>> > >>>>> >> >
>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >> _______________________________________________
>> > >>>>> >> Opensim-dev mailing list
>> > >>>>> >> Opensim-dev at opensimulator.org
>> > >>>>> >> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> >>
>> > >>>>> >
>> > >>>>> >
>> > >>>>> >
>> > >>>>> > _______________________________________________
>> > >>>>> > Opensim-dev mailing list
>> > >>>>> > Opensim-dev at opensimulator.org
>> > >>>>> > http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>> _______________________________________________
>> > >>>>> Opensim-dev mailing list
>> > >>>>> Opensim-dev at opensimulator.org
>> > >>>>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>>
>> > >>>>
>> > >>>>
>> > >>>> _______________________________________________
>> > >>>> Opensim-dev mailing list
>> > >>>> Opensim-dev at opensimulator.org
>> > >>>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>>
>> > >>>>
>> > >>>
>> > >>> _______________________________________________
>> > >>> Opensim-dev mailing list
>> > >>> Opensim-dev at opensimulator.org
>> > >>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>>
>> > >>>
>> > >>
>> > >> _______________________________________________
>> > >> Opensim-dev mailing list
>> > >> Opensim-dev at opensimulator.org
>> > >> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >>
>> > >>
>> > > _______________________________________________
>> > > Opensim-dev mailing list
>> > > Opensim-dev at opensimulator.org
>> > > http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>> > >
>> > >
>> >
>> >
>> >
>> > _______________________________________________
>> > Opensim-dev mailing list
>> > Opensim-dev at opensimulator.org
>> > http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>>
>> _______________________________________________
>> Opensim-dev mailing list
>> Opensim-dev at opensimulator.org
>> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20151113/7f626442/attachment-0001.html>
More information about the Opensim-dev
mailing list