Re: [PATCH] Win32: simplify loading of DLL functions

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Sep 20, 2017, 21:16 UTC
Message-ID: <alpine.DEB.2.21.1.1709202235300.219280@virtualbox>
In-Reply-To: <20170919180742.GC75068@aiede.mtv.corp.google.com>

Hi Jonathan,

On Tue, 19 Sep 2017, Jonathan Nieder wrote:

Show 27 quoted lines

> Johannes Schindelin wrote:
> 
> > Dynamic loading of DLL functions is duplicated in several places in Git
> > for Windows' source code.
> >
> > This patch adds a pair of macros to simplify the process: the
> > DECLARE_PROC_ADDR(<dll>, <return-type>, <function-name>,
> > ...<function-parameter-types>...) macro to be used at the beginning of a
> > code block, and the INIT_PROC_ADDR(<function-name>) macro to call before
> > using the declared function. The return value of the INIT_PROC_ADDR()
> > call has to be checked; If it is NULL, the function was not found in the
> > specified DLL.
> >
> > Example:
> >
> >         DECLARE_PROC_ADDR(kernel32.dll, BOOL, CreateHardLinkW,
> >                           LPCWSTR, LPCWSTR, LPSECURITY_ATTRIBUTES);
> >
> >         if (!INIT_PROC_ADDR(CreateHardLinkW))
> >                 return error("Could not find CreateHardLinkW() function";
> >
> > 	if (!CreateHardLinkW(source, target, NULL))
> > 		return error("could not create hardlink from %S to %S",
> > 			     source, target);
> > 	return 0;
> 
> nit: whitespace is a bit strange here (mixture of tabs and spaces).

Heh, I never thought that this would be a problem in *commit messages*. TBH I simply copy-edited the code from Git for Windows' source code.

> Could this example go near the top of the header instead?  That way,
> it's easier for people reading the header to see how to use it.

Funny, I am *so* used to examples being at the very end, from tutorials to man pages.

If my experience is any indication, I would rather keep this order.

Show 5 quoted lines

> > Signed-off-by: Karsten Blees <blees@dcon.de>
> > Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> > ---
> 
> Just curious: what was Karsten's contribution?

Essentially all of the code. I just moved it around, separated it out from the patch introducing it, used it elsewhere (e.g. in compat/poll/, not yet contributed to git/git) and replaced the LoadLibrary() call by LoadLibraryEx() for more precise control over the operation.

> (I ask mostly because I'm interested in kinds of collaboration git
> metadata is failing to capture correctly --- e.g. pair programming.)

Well, then I have this challenge for you: make use of the history of Git for Windows' master branch. It is rebased to git/git's maint (or for .0 versions, master) using the "merging rebase" strategy (i.e. first merging the to-be-rebased history using the "ours" strategy, then applying the patches on top, so that the previous commit history is part of the new history, yet there are still rebased patches ready to be prepared for submission to git/git). This strategy means that multiple versions of the same patches exist.

Show 11 quoted lines

> >         So far, there are no users (except in Git for Windows). Ben
> >         promised to make use of it in his fsmonitor patch series.
> >
> >  compat/win32/lazyload.h | 44 ++++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 44 insertions(+)
> >  create mode 100644 compat/win32/lazyload.h
> 
> Are any of the Git for Windows users something that could go upstream
> along with this patch?  That would help illustrate what a good caller
> looks like, which should help with reviewing future patches that use
> this code.

I do not currently have the time to do that, that's why I did not accompany the patch by any user.

However, having said that, Ben's patch series will make for an *excellent* user, fulfilling your wish.

Show 21 quoted lines

> > --- /dev/null
> > +++ b/compat/win32/lazyload.h
> > @@ -0,0 +1,44 @@
> [...]
> > +/* Declares a function to be loaded dynamically from a DLL. */
> > +#define DECLARE_PROC_ADDR(dll, rettype, function, ...) \
> > +	static struct proc_addr proc_addr_##function = \
> > +	{ #dll, #function, NULL, 0 }; \
> > +	static rettype (WINAPI *function)(__VA_ARGS__)
> > +
> > +/*
> > + * Loads a function from a DLL (once-only).
> > + * Returns non-NULL function pointer on success.
> > + * Returns NULL + errno == ENOSYS on failure.
> > + */
> > +#define INIT_PROC_ADDR(function) \
> > +	(function = get_proc_addr(&proc_addr_##function))
> 
> Probably worth mentioning in the doc comment that this is not thread
> safe, so a caller that wants to lazy-init in a threaded context is
> responsible for doing their own locking.

True.

Show 20 quoted lines

> > +static inline void *get_proc_addr(struct proc_addr *proc)
> > +{
> > +	/* only do this once */
> > +	if (!proc->initialized) {
> > +		HANDLE hnd;
> > +		proc->initialized = 1;
> > +		hnd = LoadLibraryExA(proc->dll, NULL,
> > +				     LOAD_LIBRARY_SEARCH_SYSTEM32);
> > +		if (hnd)
> > +			proc->pfunction = GetProcAddress(hnd, proc->function);
> > +	}
> > +	/* set ENOSYS if DLL or function was not found */
> > +	if (!proc->pfunction)
> > +		errno = ENOSYS;
> > +	return proc->pfunction;
> > +}
> 
> strerror(ENOSYS) is "Function not implemented".  Cute.
> 
> Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>

Okay, I'll add that for v2. Will wait a couple of days in case more stuff crops up.

Ciao, Dscho