Hi Daniel! Just simply took a glance at the code, I'm curious why you don't just call scm_init_guile() in fuse operation init function? This should let you call scm_init_guile() just once.
Best regards. On Mon, Mar 3, 2025 at 6:41 PM Daniel Littlewood via General Guile related discussions <guile-user@gnu.org> wrote: > Hi guile-users, > > I am working on a guile program to interact with FUSE. The code is at > https://gitlab.com/danielittlewood0/diffuse. I have managed to get it to > basically work, but there are some things I don't understand. It > consists of essentially two files at the moment: > > * other.scm, which implements the actual logic of the filesystem in > guile. > * hello.c, which does the plumbing into the FUSE library. > > I managed to get it to work by doing the following in each of the FUSE > handlers: > > scm_init_guile() > scm_c_primitive_load("/full/path/to/other.scm") > scm_call_1(scm_variable_ref(scm_c_lookup("whatever-function-I-want"))) > > That worked fine, but I was a bit confused about having to call > scm_init_guile so many times. I expected to just call it once in main > and then never again. If I do that, i.e. just move scm_init_guile to > main and remove all the other callers (rest of the code unchanged) I get > a segfault. > > I was also surprised that I had to give the absolute file path of my > guile script to scm_c_primitive_load. If I called it from main then I > didn't have to do that. If I call scm_c_primitive_load from main then I > can give a relative path, but from e.g. hello_read I get a "No such file > or directory" error and have to provide the absolute path. > > I think this is something peculiar to how FUSE is set up, but I don't > know what exactly. I don't actually call the handlers, I pass the > function pointers to the library and the library calls them for me. For > instance if I try to print the current working directory from within one > of the handlers, I get / - which probably refers to the root of the FUSE > filesystem tree. I managed to avoid giving the absolute path by calling > scm_init_guile() and then scm_c_primitive_load("other.scm") from main. > This works (with the relative path). What's surprising to me is: > > * I still have to call scm_init_guile before every use of guile within > the current function, even if I called it in main. > * When I call scm_init_guile a second time, I don't have to call > scm_primitive_load a second time. I can just call scm_variable_ref > immediately. > > This is a bit confusing to me. I could understand if I had to load the > code every time I call scm_init_guile (although then I'd want to reduce > the boilerplate) and I could understand calling scm_init_guile and > scm_c_primitive_load only once, but the arrangement above doesn't make > sense to me. I feel like I must be missing something. > > Note that this is the first project I've really written in either C or > guile, so I might be missing something obvious. > > Best wishes, > Dan >
