Usually when loading fails but the file exists, it means that dlopen tried loading the file but it depended on other libraries that couldn’t be found. But I’m not sure why that would happen, unless your libblas.so.3 is different from the one on my system. You can run ldd on the copy of libblas.so.3 in your distribution’s lib directory (with LD_LIBRARY_PATH unset) to see if it complains about any dependencies not being found.

@rajan.yadav I think, you are close to get it working.

Just checking: To load libblas the dependencies (in your case libgfortran, and liblapack) needs to be available too. That is, they also need to be copied into the lib folder.

Second: If loading libblas still fails, then you need to check the actual paths inside the shared libraries to see if the path is as you expect. I believe ldd is the command used on Linux:

ldd libblas.so.3

I just published a package of my random utility libraries, and I added a variant of cond, but with a different syntax from sorawee’s version above. See cond+ in scramble/cond from the scramble-lib package.

Hi, I’d like to learn logic programming. I plan to use The Reasoned Schemer 2ed because it looks very interesting. Question: If I want to run the examples in that book, which Racket library should I use? Thanks.

Also very appreciate if other books recommendations.

https://www.monolune.com/using-racket-for-the-reasoned-schemer/

https://docs.racket-lang.org/minikanren/index.html

Thanks.

The (require minikanren) seemly only works in 1ed, not in 2ed. For example, 1ed and minikanren (run∗ (q) #u) 2ed, (run∗ q #u) ; no parentheses around q And no build-in #s and #u.

I have the first edition. Does the book have anything to say on how to install minikanren?

No.

I think there were some comments about 2nd Ed below the blog post @soegaard2 linked

An alternative is to use the same style of development, that Byrds uses. Simply save “trs2-impl.scm” next to your file. Then use (load "trs2-impl.scm") and enter everything in the repl.

Yes, but they use Scheme… Anyway it would be nice if have Racket version.

hayesall has a fix: https://github.com/TheReasonedSchemer2ndEd/CodeFromTheReasonedSchemer2ndEd/issues/1

@soegaard2 It works. Thanks!

All needed libs are in /me/mypkg/lib/ . LD_LIBRARY_PATH when set manually, just points to that dir.

There could be absolute paths inside libblas. It’s worth checking.

PS. Even in William Byrd’s solution, #u and #s doesn’t work out of the box and monolune’s readtable hack doesn’t work for me. Although it is not a big problem because we can use succeed and fail instead.

$ ldd mypkg/lib/libblas.so.3             linux-vdso.so.1 (0x00007ffe205bd000)     libgfortran.so.4 => not found     libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f447fc69000)     libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f447f878000)     /lib64/ld-linux-x86-64.so.2 (0x00007f44805c9000) $ ls mypkg/lib/ libblas.so libblas.so.3 libffi.so.6 libgfortran.so.3 libgfortran.so.4 liblapack.so.3 plt

Its not finding libgfortran, but its there

Need to keep digging into it :slightly_smiling_face:

To get #u and #s working the reader needs to change. I think, #u and #s are used in the book only.

Hmm.

So I need to build a version of these libs that does look into current directory

Or inject rpath into these libs using patchelf or similar tool

Let’s try to change the order in which flomat loads the libraries.

Right?

If libgfortran is loaded first, perhaps libblas will load correctly?

Okay yes. I can try that

The code that loads libblas on Linux is:

So try moving the (define gfortran-lib ...) above (define cblas-lib ...).

I’ll play with it in a bit. Currently AFK. Thanks

Hi! Has anyone used the sdl2 package, particularly in the “pretty” mode? I can’t figure out how to specify a Rect struct with it…

Looks like changing the order worked as you suggested @soegaard2

hello ,I am trying build the racket from source code in https://github.com/racket/racket ,I executed build.bat in C:\Users\xx\Desktop\racket-master\racket\src\worksp ,and there is no error. but it get a runtime error like this： can anyone help?

It say"Failure: cannot load DLL", but obvious there has the correctly libracketxxxxxxx.dll

I find “libracketxxxxxxx.dll” in github and find nothing

that’s the last message in building process

Can you check “libracketxxxxxxx.dll” with something like “depends.exe” to see whether it depends on some other DLL that is missing? I could check if you upload “libracketxxxxxxx.dll” somewhere.

Is there a way to deactivate all contract checking in a part of the code? Some background: I had replaced car and cdr with first and rest in a time-critical function and noticed that this doubled the runtime of my program, not just the function. I reverted the change, but I’m also still wondering if I could save yet more time by deactivating contracts on vector-ref and vector-set!, for example. According to raco profile, the function at the moment contributes about 30 % of the program’s runtime.

You can use (#%declare #:unsafe)

That is, of course, unsafe

That looks right for “RacketCGC.exe”. What are the dependencies of “libracketxxxxxxx.dll”?

I re-used the x64 Native Tools Command Prompt for vs2019 to compile with administrator privileges and did not solve any problems

As I understand the documentation, this affects the whole module, doesn’t it?

Yes

If you want to do it for a smaller region I would make that into a submodule

So if I wanted to use it only on some specific part of the code, I’d have to move this code to a distinct module.

Ok, thought that. :slightly_smiling_face:

You can also just use individual unsafe operations

Anyway, I added (#%declare #:unsafe) for the whole module and found it hardly made a difference. So it doesn’t make any (more) difference, which is an interesting insight as well.

That does look likely to be the problem. Does turning down the arrow there show which import is missing?

> You can also just use individual unsafe operations You mean from https://docs.racket-lang.org/reference/unsafe.html ? Are there more than these on the page? Just asking.

I had seen the page at some point, but hadn’t seen that there were so many of these functions. Good. :+1:

I don’t know how to display all the problems at once, I found this

@rajan.yadav That’s great. Can you paste the exact version of the code that worked for you?

Should I download an ext-ms-win-she1132-shellcom-l1–1–0.dll To try to solve this problem?

Probably not. When I try (what appears to be) the same dependency walker that you’re using, I see the same result, so I guess it’s not the real problem.

Just to make sure, your get the same error dialog if you run racket\racketcgc.exe directly, right?

yes

Oh — looking again at your original dependency list for “libracketxxxxxxx.dll”, I see “libffi.dll”, which isn’t in my build’s dependencies.

you mean this one?

Yes.

I’m not sure why libffi is dynamically linked instead of statically linked, but if you go to “racket\lib” and run “..\racketcgc.exe” from there, does it start? (By going to “racket\lib”, “libffi.dll” will be in the current directory.)

I cannot input anything

Ok - it’s going to take a very long time, since “.rkt” files are not yet compiled, but that at least started up.

I can look more into how libffi got linked as a DLL. Can you tell me how you started the Racket build after checking out the repo?

Ah, sorry, when the problem happened for the first time, I set the environment variable

I directly executed C:\Users\mingy\Desktop\racket-master\racket\src\worksp\build.bat in x64 Native Tools Command Prompt for vs 2019. This is because I used to compile version 7.5 in the same way.

This is because I used the same way when compiling racket7.5 a long time ago

The result of executing ..\racketcgc.exe in the lib directory just now is that it exited directly. I went back to the directory where racketcgc.exe was located and double-clicked it. The previous problem still exists.

Maybe something has changed (maybe as part of the libffi upgrade) so that runing just “build.bat” doesn’t work right, but it’s supposed to build BC — but with a “BC” suffix. While I look into what happens when you run “build.bat” directly, you might have more success using just nmake (to get CS build) or nmake win-bc (to get a BC build) in the repo top-level directory.

Should I re-extract the source code to maintain a clean environment, and then execute make?

Yes. If you used git, then I’d go with git clean -d -x -f, but re-extracting sounds right if you downloaded an archive

I think you’re still going to run into trouble, though, since “build.bat” still works for me.

:sob:

I used the windows insider version. Could this be the reason for our different results?

That doesn’t seem too likely.

Oh, I bet it’s VS 2019 versus 2017.

I haven’t yet tried 2019

yes

I noticed that loaded is null

If you open “racket\src\worksp\racket\racketX.sln” in Visual Studio, right-click on the “libffiX” subproject and select “Properties”, what does it show as the “Confoguration type”? Mine says “Static library (.lib)”

do you mean this?

Yes, thanks

How about libracketX -> Properties -> Linker -> Command Line?

/OUT:"......\lib\libracketxxxxxxx.dll" /MANIFEST /PDB:"......\lib\libracketxxxxxxx.pdb" /DYNAMICBASE:NO “WS2_32.lib” “Shell32.lib” “User32.lib” “Winmm.lib” “kernel32.lib” “user32.lib” “gdi32.lib” “winspool.lib” “comdlg32.lib” “advapi32.lib” “shell32.lib” “ole32.lib” “oleaut32.lib” “uuid.lib” “odbc32.lib” “odbccp32.lib” /IMPLIB:"......\lib\msvc\libracketxxxxxxx.lib" /DEBUG /DLL /BASE:"0x10400000" /MACHINE:X64 /PGD:"......\lib\libracketxxxxxxx.pgd" /MANIFESTUAC:"level=’asInvoker’ uiAccess=’false’" /ManifestFile:"x64\SDebug\libracketxxxxxxx.dll.intermediate.manifest" /LTCGOUT:"x64\SDebug\libracketxxxxxxx.iobj" /ERRORREPORT:PROMPT /NOLOGO /TLBID:1

Sorry, I thought the text mode would make it easier for you to search

I expected to see “libffi.lib” in that list. I guess something has changed in the way Visual Studio treats project dependencies, so I’ll probably have to install 2019 or find an installation to sort it out.

Meanwhile, probably you can make progress by temporarily adding racket\lib to your PATH so that “libffi.dll” will be found.

I read this, and its default state is like this

Do you mean this? what name should I type?

or like this?

but it did not work。。

Did you start a fresh command prompt after changing that?

I did it by double clicking it

Just to make sure, if you start a new command prompt and use “set” to show environment variables, the new path is in PATH? And running racketcgc in that prompt still fails?

yes

I’m stumped for now, and I’ll probably just have to create a VS 2019 installation.

I noticed that the loading of libmzgcxxxxxxx.dll was successful,

But the second one failed

Sorry I have to go to bed first, I’m at UTC+08:00, and it’s been a night。

晚安. I’m installing VS 2019, so maybe things will just work tomorrow.

No… building with VS 2019 worked for me.

I’m configuring my environment by running the “racket\src\worksp\msvcpep.bat” script. Maybe there’s a difference in how our VS environments are set up?

Using x64 Native Tools Command Prompt made no difference, either

Maybe it really is something about the OS, somehow. I’m out of other ideas.

Ok, one more idea: maybe there’s something in the output of build.bat for the msbuild where it’s building libffi. Here’s the output I get.

[(unix)                                                            ; Note: The library names are different on Debian, Ubuntu and Arch.                             (define uname (string-downcase (system-type 'machine)))                                   (define dist (cond [(regexp-match "arch"    uname) 'arch]                                          [(regexp-match "debian"   uname) 'debian]                                         [(regexp-match "ubuntu"   uname) 'ubuntu]                                         [(regexp-match #px"fc\\d\\d" uname) 'fedora]                                         [else                'other]))                                                                                            (define gfortran-lib (case dist                                                          [(fedora) (ffi-lib "libgfortran" '("5" #f))]                                        [else   (ffi-lib "libgfortran" '("3" #f))]))                            ; The lib order is important here.                                              (define cblas-lib  (case dist                                                          [(debian) (ffi-lib "libblas" '("3" #f))]                                          [(arch)  (ffi-lib "libcblas" '("3" #f))]                                          [(ubuntu) (ffi-lib "libblas" '("3" #f))]                                          [(fedora) (ffi-lib "libcblas" '("3" #f))]                                          [(other) (ffi-lib "libblas" '("3" #f))]))                                                                                            (define quadmath-lib (ffi-lib "libquadmath" '("0" #f)))                                   (define lapack-lib  (ffi-lib "liblapack"  '("3" #f)))                                   (values cblas-lib lapack-lib)] Here it is

I don’t know why this happens, but it may be related on what mouse messages are or are not sent to the widget — I fixed similar issues on Windows. It would be good if you can write a short program that shows this behavior, so we can test if this is a problem on MacOS only or it affects other platforms…

I’ll make a small example example later. FWIW I am using an Apple Magic Trackpad 2, so it is possible the messages differ from a normal mouse.