Unable to correctly identify a type of Windows static library name, such as foo.lib

[spritetong]

I have made a patch to fix it.

fn is_static_available(name: &str, system_roots: &[PathBuf], dirs: &[PathBuf]) -> bool {
    let libname = format!("lib{}.a", name);
    let libname_win = if cfg!(windows) {
        format!("{}.lib", name)
    } else {
        Default::default()
    };

    dirs.iter().any(|dir| {
        !system_roots.iter().any(|sys| dir.starts_with(sys))
            && (dir.join(&libname).exists()
                || (!libname_win.is_empty() && dir.join(&libname_win).exists()))
    })
}

diff --git a/src/lib.rs b/src/lib.rs
index 2c690fa..a5df218 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -832,9 +832,16 @@ fn envify(name: &str) -> String {
 /// System libraries should only be linked dynamically
 fn is_static_available(name: &str, system_roots: &[PathBuf], dirs: &[PathBuf]) -> bool {
     let libname = format!("lib{}.a", name);
+    let libname_win = if cfg!(windows) {
+        format!("{}.lib", name)
+    } else {
+        Default::default()
+    };
 
     dirs.iter().any(|dir| {
-        !system_roots.iter().any(|sys| dir.starts_with(sys)) && dir.join(&libname).exists()
+        !system_roots.iter().any(|sys| dir.starts_with(sys))
+            && (dir.join(&libname).exists()
+                || (!libname_win.is_empty() && dir.join(&libname_win).exists()))
     })
 }
 

[sdroege]

Unfortunately it's not that easy. .lib files are also DLL import libraries.

To know whether it's one or another you actually need to parse the file.

[spritetong]

Yes, there's no simple way to automatically tell a .lib file is an archive or a DLL import library without parsing its content.
But the problem is, if I'm sure this is a static library, Config::probe() always treats it as a shared library.

If I call,
pkg_config::Config::new().statik(true).probe("foo").unwrap();
then
if statik && is_static_available(val, &system_roots, &dirs) {
failed.

Another solution is to add a special option to Config that forces it to link the static library, like

pkg_config::Config::new().force_static(true).probe("foo").unwrap();

[ChrisDenton]

Hmm... on Windows both static .lib files and DLL import libraries are really static libraries. You can even mix the two in the same .lib. So I'm uncertain why the distinction matters here? Does something fail in some way?

[spritetong]

Hmm..., I mean, for a static library named foo.lib, is_static_available() always returns false, even if statik(true) is set.
There's no way to force is_static_available() to return true.

[ChrisDenton]

Right but if we look at the code that uses it, there doesn't seem to much difference:

pkg-config-rs/src/lib.rs

Lines 718 to 726 in d039d32

	if statik && is_static_available(val, &system_roots, &dirs) {
	let meta = format!("rustc-link-lib=static={}", val);
	config.print_metadata(&meta);
	} else {
	let meta = format!("rustc-link-lib={}", val);
	config.print_metadata(&meta);
	}
	self.libs.push(val.to_string());

The only difference between the two branches is =static is added in one case and in the other it's left up to rustc to decide.

[spritetong]

For a static library, if it runs to the branch else { ... } , MSVC linker will fail.

[ChrisDenton]

I'm confused as to why that would be as rustc treats them the same way (see here and here):

    fn link_dylib(&mut self, lib: &str, verbatim: bool, _as_needed: bool) {
        self.cmd.arg(format!("{}{}", lib, if verbatim { "" } else { ".lib" }));
    }
    fn link_staticlib(&mut self, lib: &str, verbatim: bool) {
        self.cmd.arg(format!("{}{}", lib, if verbatim { "" } else { ".lib" }));
    }

I think it would help to have a minimal example to test this. It could be a problem in rust or cargo itself.