0ad/source/lib/res/file/vfs_mount.cpp

#include "precompiled.h"

#include "../res.h"
#include "vfs_mount.h"
#include "vfs_path.h"
#include "vfs_tree.h"
#include "zip.h"
#include "sysdep/dir_watch.h"

#include <deque>
#include <list>

struct Stats
{
	size_t mounted_dirs;
	size_t mounted_archives;
	size_t files_examined;
	size_t files_updated;

	void dump()
	{
		debug_printf("VFS stats\n");
		debug_printf("  dirs=%u archives=%u\n", mounted_dirs, mounted_archives);
		debug_printf("  files updated=%u files examined=%u\n", files_updated, files_examined);
	}
};
static Stats stats;

void vfs_dump_stats()
{
	stats.dump();
}


// location of a file: either archive or a real directory.
// not many instances => don't worry about efficiency.
struct Mount
{
	// mounting into this VFS directory;
	// must end in '/' (unless if root td, i.e. "")
	std::string V_mount_point;

	// real directory being mounted.
	// if this Mount represents an archive, this is the real directory
	// containing the Zip file (required so that this Mount is unmounted).
	std::string P_name;

	Handle archive;

	uint pri;

	// see enum VfsMountFlags
	uint flags;

	MountType type;

	Mount(const char* V_mount_point_, const char* P_name_, Handle archive_, uint flags_, uint pri_)
		: V_mount_point(V_mount_point_), P_name(P_name_)
	{
		archive = archive_;
		flags = flags_;
		pri = pri_;

		if(archive > 0)
		{
			h_add_ref(archive);
			type = MT_ARCHIVE;
		}
		else
			type = MT_FILE;
	}

	~Mount()
	{
		if(archive > 0)	// avoid h_mgr warning
			zip_archive_close(archive);
	}

	Mount& operator=(const Mount& rhs)
	{
		V_mount_point = rhs.V_mount_point;
		P_name        = rhs.P_name;
		archive       = rhs.archive;
		pri           = rhs.pri;
		flags         = rhs.flags;
		type          = rhs.type;

		if(archive > 0)	// avoid h_mgr warning
			h_add_ref(archive);

		return *this;
	}

	struct equal_to : public std::binary_function<Mount, const char*, bool>
	{
		bool operator()(const Mount& m, const char* P_name) const
		{
			return (m.P_name == P_name);
		}
	};

private:
	Mount();
};



char mount_get_type(const Mount* m)
{
	switch(m->type)
	{
	case MT_ARCHIVE:
		return 'A';
	case MT_FILE:
		return 'F';
	default:
		return '?';
	}
}


bool mount_should_replace(const Mount* m_old, const Mount* m_new, bool files_are_identical)
{
	// need to "replace" if not yet associated with a Mount
	if(!m_old)
		return true;

	// keep old if new priority is lower.
	if(m_new->pri < m_old->pri)
		return false;

	// since priority is not less, we really ought to always go with m_new.
	// however, there is one special case we handle for performance reasons:
	// if the file contents are the same, prefer the more efficient source.
	// note that priority doesn't automatically take care of this,
	// especially if set incorrectly.
	if(files_are_identical && m_old->type > m_new->type)
		return false;

	return true;
}


///////////////////////////////////////////////////////////////////////////////
//
// populate the directory being mounted with files from real subdirectories
// and archives.
//
///////////////////////////////////////////////////////////////////////////////

static const Mount& add_mount(const char* V_mount_point, const char* P_real_path, Handle archive,
	uint flags, uint pri);

// passed through dirent_cb's zip_enum to zip_cb
struct ZipCBParams
{
	// tree directory into which we are adding the archive's files
	TDir* const td;

	// archive's location; assigned to all files added from here
	const Mount* const m;

	// storage for directory lookup optimization (see below).
	// held across one zip_enum's zip_cb calls.
	char last_path[VFS_MAX_PATH];
	size_t last_path_len;
	TDir* last_td;

	ZipCBParams(TDir* dir_, const Mount* loc_)
		: td(dir_), m(loc_)
	{
		last_path[0] = '\0';
		last_path_len = 0;
		last_td = 0;
	}
	// no copy ctor because some members are const
private:
	ZipCBParams& operator=(const ZipCBParams&);
};

// called by add_ent's zip_enum for each file in the archive.
// we get the full path, since that's what is stored in Zip archives.
//
// [total time 21ms, with ~2000 file's (includes add_file cost)]
static int zip_cb(const char* path, const struct stat* s, uintptr_t user)
{
	CHECK_PATH(path);

	ZipCBParams* params = (ZipCBParams*)user;
	TDir* td             = params->td;
	const Mount* m        = params->m;
	char* last_path       = params->last_path;
	size_t& last_path_len = params->last_path_len;
	TDir*& last_td       = params->last_td;

	// extract file name (needed for add_file)
	const char* fn = path;
	const char* slash = strrchr(path, '/');
	if(slash)
		fn = slash+1;
	// else: there is no path - it's in the archive's root td.

	// into which directory should the file be inserted?
	// naive approach: tree_lookup_dir the path (slow!)
	// optimization: store the last file's path; if it's the same,
	//   use the directory we looked up last time (much faster!)
	const size_t path_len = fn-path;
	// .. same as last time
	if(last_td && path_len == last_path_len &&
		strnicmp(path, last_path, path_len) == 0)
		td = last_td;
	// .. last != current: need to do lookup
	else
	{
		vfs_path_copy(last_path, path);
		last_path_len = path_len;
		last_path[last_path_len] = '\0';
		// strip filename (tree_lookup_dir requirement)

		CHECK_ERR(tree_lookup_dir(last_path, &td, LF_CREATE_MISSING|LF_START_DIR));
		// we have to create them if missing, since we can't rely on the
		// archiver placing directories before subdirs or files that
		// reference them (WinZip doesn't always).
		// we also need to start at the mount point (td).

		last_td = td;
	}

	return tree_add_file(td, fn, m, s->st_size, s->st_mtime);
}


static bool archive_less(Handle hza1, Handle hza2)
{
	const char* fn1 = h_filename(hza1);
	const char* fn2 = h_filename(hza2);
	return strcmp(fn1, fn2) < 0;
}

typedef std::vector<Handle> Archives;
typedef Archives::const_iterator ArchiveCIt;

static int enqueue_archive(const char* name, const char* P_archive_dir, Archives* archives)
{
	// caller wants us to check if this is a Zip file and if so, append it to
	// a list. this is only done in the mounted directory itself, not its
	// subdirectories! see mount_dir_tree.
	// the archives will be mounted after regular directory mounts are done.
	if(archives)
	{
		// get complete path for zip_archive_open.
		// this doesn't (need to) work for subdirectories of the mounted td!
		// we can't use mount_get_path because we don't have the VFS path.
		char P_path[PATH_MAX];
		vfs_path_append(P_path, P_archive_dir, name);

		// just open the Zip file and see if it's valid. we don't bother
		// checking the extension because archives won't necessarily be
		// called .zip (e.g. Quake III .pk3).
		Handle archive = zip_archive_open(P_path);
		if(archive > 0)
		{
			archives->push_back(archive);

			// avoid also adding the Zip file itself to <td>.
			return 0;
		}
	}

	return 1;
}

static int mount_archive(TDir* td, const Mount& m)
{
	stats.mounted_archives++;

	ZipCBParams params(td, &m);
	zip_enum(m.archive, zip_cb, (uintptr_t)&params);
	return 0;
}

static int mount_archives(TDir* td, Archives* archives, const Mount* mount)
{
	// VFS_MOUNT_ARCHIVES flag wasn't set, or no archives present
	if(archives->empty())
		return 0;

	std::sort(archives->begin(), archives->end(), archive_less);

	for(ArchiveCIt it = archives->begin(); it != archives->end(); ++it)
	{
		Handle hza = *it;

		// add this archive to the mount list (address is guaranteed to
		// remain valid).
		const Mount& m = add_mount(mount->V_mount_point.c_str(), mount->P_name.c_str(), hza, 0, 0);

		mount_archive(td, m);
	}

	return 0;
}


//-----------------------------------------------------------------------------

struct TDirAndPath
{
	TDir* const td;
	const std::string path;

	TDirAndPath(TDir* d, const char* p)
		: td(d), path(p)
	{
	}

	// no copy ctor because some members are const
private:
	TDirAndPath& operator=(const TDirAndPath&);
};

typedef std::deque<TDirAndPath> DirQueue;



static int enqueue_dir(TDir* parent_td, const char* name,
	const char* P_parent_path, DirQueue* dir_queue)
{
	// caller doesn't want us to enqueue subdirectories; bail.
	if(!dir_queue)
		return 0;

	// skip versioning system directories - this avoids cluttering the
	// VFS with hundreds of irrelevant files.
	// we don't do this for Zip files because it's harder (we'd have to
	// strstr the entire path) and it is assumed the Zip file builder
	// will take care of it.
	if(!strcmp(name, "CVS") || !strcmp(name, ".svn"))
		return 0;

	// prepend parent path to get complete pathname.
	char P_path[PATH_MAX];
	CHECK_ERR(vfs_path_append(P_path, P_parent_path, name));

	// create subdirectory..
	TDir* td;
	CHECK_ERR(tree_add_dir(parent_td, name, &td));
	// .. and add it to the list of directories to visit.
	dir_queue->push_back(TDirAndPath(td, const_cast<const char*>(P_path)));
	return 0;
}




// called by TDir::addR's file_enum for each entry in a real directory.
//
// if called for a real directory, it is added to VFS.
// else if called for a loose file that is a valid archive (*),
//   it is mounted (all of its files are added)
// else the file is added to VFS.
//
// * we only perform this check in the directory being mounted,
// i.e. passed in by tree_add_dir. to determine if a file is an archive,
// we have to open it and read the header, which is slow.
// can't just check extension, because it might not be .zip (e.g. Quake3 .pk3).

//
// td - tree td into which the dirent is to be added
// m - real td's location; assigned to all files added from this mounting
// archives - if the dirent is an archive, its Mount is added here.

static int add_ent(TDir* td, DirEnt* ent, const char* P_parent_path, const Mount* m,
	DirQueue* dir_queue, Archives* archives)
{
	const char* name = ent->name;

	// it's a directory entry.
	if(DIRENT_IS_DIR(ent))
		return enqueue_dir(td, name, P_parent_path, dir_queue);
	// else: it's a file (dir_next_ent discards everything except for
	// file and subdirectory entries).

	if(enqueue_archive(name, m->P_name.c_str(), archives) == 0)
		return 0;

	// it's a regular data file; add it to the directory.
	return tree_add_file(td, name, m, ent->size, ent->mtime);
}


// note: full path is needed for the dir watch.
static int populate_dir(TDir* td, const char* P_path, const Mount* m,
	DirQueue* dir_queue, Archives* archives, int flags)
{
	int err;

	RealDir* rd = tree_get_real_dir(td);
	WARN_ERR(mount_attach_real_dir(rd, P_path, m, flags));

	DirIterator d;
	RETURN_ERR(dir_open(P_path, &d));

	DirEnt ent;
	for(;;)
	{
		// don't RETURN_ERR since we need to close d.
		err = dir_next_ent(&d, &ent);
		if(err != 0)
			break;

		err = add_ent(td, &ent, P_path, m, dir_queue, archives);
		WARN_ERR(err);
	}

	WARN_ERR(dir_close(&d));
	return 0;
}


// actually mount the specified entry. split out of vfs_mount,
// because when invalidating (reloading) the VFS, we need to
// be able to mount without changing the mount list.
// add all loose files and subdirectories (recursive).
// also mounts all archives in P_real_path and adds to archives.

// add the contents of directory <p_path> to this TDir,
// marking the files' locations as <m>. flags: see VfsMountFlags.
//
// note: we are only able to add archives found in the root directory,
// due to dirent_cb implementation. that's ok - we don't want to check
// every single file to see if it's an archive (slow!).
static int mount_dir_tree(TDir* td, const Mount& m)
{
	int err = 0;

	// add_ent fills these queues with dirs/archives if the corresponding
	// flags are set.
	DirQueue dir_queue;	// don't preallocate (not supported by TDirAndPath)
	Archives archives;
	archives.reserve(8); // preallocate for efficiency.

	// instead of propagating flags down to add_dir, prevent recursing
	// and adding archives by setting the destination pointers to 0 (easier).
	DirQueue* const pdir_queue = (m.flags & VFS_MOUNT_RECURSIVE)? &dir_queue : 0;
	Archives* parchives = (m.flags & VFS_MOUNT_ARCHIVES)? &archives : 0;

	// kickoff (less efficient than goto, but c_str reference requires
	// pop to come at end of loop => this is easiest)
	dir_queue.push_back(TDirAndPath(td, m.P_name.c_str()));

	do
	{
		TDir* const td     = dir_queue.front().td;
		const char* P_path = dir_queue.front().path.c_str();

		int ret = populate_dir(td, P_path, &m, pdir_queue, parchives, m.flags);
		if(!err)
			err = ret;

		// prevent searching for archives in subdirectories (slow!). this
		// is currently required by the implementation anyway.
		parchives = 0;

		dir_queue.pop_front();
		// pop at end of loop, because we hold a c_str() reference.
	}
	while(!dir_queue.empty());

	// do not pass parchives because that has been set to 0!
	mount_archives(td, &archives, &m);

	return 0;
}







// the VFS stores the location (archive or directory) of each file;
// this allows multiple search paths without having to check each one
// when opening a file (slow).
//
// one Mount is allocated for each archive or directory mounted.
// therefore, files only /point/ to a (possibly shared) Mount.
// if a file's location changes (e.g. after mounting a higher-priority
// directory), the VFS entry will point to the new Mount; the priority
// of both locations is unchanged.
//
// allocate via mnt_create, passing the location. do not free!
// we keep track of all Locs allocated; they are freed at exit,
// and by mnt_free_all (useful when rebuilding the VFS).
// this is much easier and safer than walking the VFS tree and
// freeing every location we find.







///////////////////////////////////////////////////////////////////////////////
//
// mount list (allows multiple mountings, e.g. for mods)
//
///////////////////////////////////////////////////////////////////////////////

// every mounting results in at least one Mount (and possibly more, e.g.
// if the directory contains Zip archives, which each get a Mount).
//
// requirements for container:
// - must not invalidate iterators after insertion!
//   (TFile holds a pointer to the Mount from which it was added)
// - must store items in order of insertion
//   xxx

typedef std::list<Mount> Mounts;
typedef Mounts::iterator MountIt;
static Mounts mounts;


static const Mount& add_mount(const char* V_mount_point, const char* P_real_path, Handle hza,
	uint flags, uint pri)
{
	mounts.push_back(Mount(V_mount_point, P_real_path, hza, flags, pri));
	return mounts.back();
}


// note: this is not a member function of Mount to avoid having to
// forward-declare mount_archive, mount_dir_tree.
static int remount(const Mount& m)
{
	TDir* td;
	CHECK_ERR(tree_lookup_dir(m.V_mount_point.c_str(), &td, LF_CREATE_MISSING));

	switch(m.type)
	{
	case MT_ARCHIVE:
		return mount_archive(td, m);
	case MT_FILE:
		return mount_dir_tree(td, m);
	default:
		debug_warn(__func__": invalid type");
		return ERR_CORRUPTED;
	}
}

static inline void unmount_all(void)
{
	mounts.clear();
}

static inline void remount_all()
{
	std::for_each(mounts.begin(), mounts.end(), remount);
}





// mount <P_real_path> into the VFS at <V_mount_point>,
//   which is created if it does not yet exist.
// files in that directory override the previous VFS contents if
//   <pri>(ority) is not lower.
// all archives in <P_real_path> are also mounted, in alphabetical order.
//
// flags determines extra actions to perform; see VfsMountFlags.
//
// P_real_path = "." or "./" isn't allowed - see implementation for rationale.
int vfs_mount(const char* V_mount_point, const char* P_real_path, int flags, uint pri)
{
	// callers have a tendency to forget required trailing '/';
	// complain if it's not there, unless path = "" (root td).
#ifndef NDEBUG
	const size_t len = strlen(V_mount_point);
	if(len && V_mount_point[len-1] != '/')
		debug_warn(__func__": path doesn't end in '/'");
#endif

	// make sure it's not already mounted, i.e. in mounts.
	// also prevents mounting a parent directory of a previously mounted
	// directory, or vice versa. example: mount $install/data and then
	// $install/data/mods/official - mods/official would also be accessible
	// from the first mount point - bad.
	// no matter if it's an archive - still shouldn't be a "subpath".
	for(MountIt it = mounts.begin(); it != mounts.end(); ++it)
	{
		if(file_is_subpath(P_real_path, it->P_name.c_str()))
		{
			debug_warn(__func__": already mounted");
			return -1;
		}
	}

	// disallow "." because "./" isn't supported on Windows.
	// it would also create a loophole for the parent td check above.
	// "./" and "/." are caught by CHECK_PATH.
	if(!strcmp(P_real_path, "."))
	{
		debug_warn(__func__": mounting . not allowed");
		return -1;
	}

	const Mount& m = add_mount(V_mount_point, P_real_path, 0, flags, pri);
	return remount(m);
}


// rebuild the VFS, i.e. re-mount everything. open files are not affected.
// necessary after loose files or directories change, so that the VFS
// "notices" the changes and updates file locations. res calls this after
// dir_watch reports changes; can also be called from the console after a
// rebuild command. there is no provision for updating single VFS dirs -
// it's not worth the trouble.
static int rebuild()
{
	tree_clear();
	tree_init();
	remount_all();
	return 0;
}


// unmount a previously mounted item, and rebuild the VFS afterwards.
int vfs_unmount(const char* P_name)
{
	// this removes all Mounts ensuing from the given mounting. their dtors
	// free all resources and there's no need to remove the files from
	// VFS (nor is this feasible), since it is completely rebuilt afterwards.

	MountIt begin = mounts.begin(), end = mounts.end();
	MountIt last = std::remove_if(begin, end,
		std::bind2nd(Mount::equal_to(), P_name));
	// none were removed - need to complain so that the caller notices.
	if(last == end)
		return ERR_PATH_NOT_FOUND;
	// trim list and actually remove 'invalidated' entries.
	mounts.erase(last, end);

	return rebuild();
}






// if <path> or its ancestors are mounted,
// return a VFS path that accesses it.
// used when receiving paths from external code.
static int make_vfs_path(const char* P_path, char* V_path)
{
	for(MountIt it = mounts.begin(); it != mounts.end(); ++it)
	{
		const Mount& m = *it;
		if(m.type != MT_FILE)
			continue;

		const char* remove = m.P_name.c_str();
		const char* replace = m.V_mount_point.c_str();

		if(path_replace(V_path, P_path, remove, replace) == 0)
			return 0;
	}

	return -1;
}


void mount_init()
{
	tree_init();
}


void mount_shutdown()
{
	tree_clear();
	unmount_all();
}






int mount_attach_real_dir(RealDir* rd, const char* P_path, const Mount* m, int flags)
{
	// more than one real dir mounted into VFS dir
	// (=> can't create files for writing here)
	if(rd->m)
		rd->m = (const Mount*)-1;
	else
		rd->m = m;

#ifndef NO_DIR_WATCH
	if(flags & VFS_MOUNT_WATCH)
	{
		char N_path[PATH_MAX];
		CHECK_ERR(file_make_full_native_path(P_path, N_path));
		CHECK_ERR(dir_add_watch(N_path, &rd->watch));
	}
#endif

	return 0;
}


void mount_detach_real_dir(RealDir* rd)
{
	rd->m = 0;

#ifndef NO_DIR_WATCH
	if(rd->watch)	// avoid dir_cancel_watch complaining
		WARN_ERR(dir_cancel_watch(rd->watch));
	rd->watch = 0;
#endif
}


int mount_populate(TDir* td, RealDir* rd)
{
	UNUSED2(td);
	UNUSED2(rd);
	return 0;
}



//-----------------------------------------------------------------------------
// hotloading
//-----------------------------------------------------------------------------

// called by vfs_reload and vfs_reload_changed_files (which will already
// have rebuilt the VFS - doing so more than once a frame is unnecessary).
static int reload_without_rebuild(const char* fn)
{
	// invalidate this file's cached blocks to make sure its contents are
	// loaded anew.
	CHECK_ERR(file_invalidate_cache(fn));

	CHECK_ERR(h_reload(fn));

	return 0;
}


// called via console command.
int vfs_reload(const char* fn)
{
	// if <fn> currently maps to an archive, the VFS must switch
	// over to using the loose file (that was presumably changed).
	CHECK_ERR(rebuild());

	return reload_without_rebuild(fn);
}




// get directory change notifications, and reload all affected files.
// must be called regularly (e.g. once a frame). this is much simpler
// than asynchronous notifications: everything would need to be thread-safe.
int vfs_reload_changed_files()
{
	// array of reloads requested this frame (see 'do we really need to
	// reload' below). go through gyrations to avoid heap allocs.
	const size_t MAX_RELOADS_PER_FRAME = 12;
	typedef char Path[VFS_MAX_PATH];
	typedef Path PathList[MAX_RELOADS_PER_FRAME];
	PathList pending_reloads;

	uint num_pending = 0;
	// process only as many notifications as we have room for; the others
	// will be handled next frame. it's not imagineable that they'll pile up.
	while(num_pending < MAX_RELOADS_PER_FRAME)
	{
		// get next notification
		char N_path[PATH_MAX];
		int ret = dir_get_changed_file(N_path);
		CHECK_ERR(ret);	// error? (doesn't cover 'none available')
		if(ret != 0)	// none available; done.
			break;

		// convert to VFS path
		char P_path[PATH_MAX];
		CHECK_ERR(file_make_full_portable_path(N_path, P_path));
		char* V_path = pending_reloads[num_pending];
		CHECK_ERR(make_vfs_path(P_path, V_path));

		// do we really need to reload? try to avoid the considerable cost of
		// rebuilding VFS and scanning all Handles.
		//
		// note: be careful to avoid 'race conditions' depending on the
		// timeframe in which notifications reach us.
		// example: editor deletes a.tga; we are notified; reload is
		// triggered but fails since the file isn't found; further
		// notifications (e.g. renamed a.tmp to a.tga) come within x [ms] and
		// are ignored due to a time limit.
		// therefore, we can only check for multiple reload requests a frame;
		// to that purpose, an array is built and duplicates ignored.
		const char* ext = strrchr(V_path, '.');
		// .. directory change notification; ignore because we get
		//    per-file notifications anyway. (note: assume no extension =>
		//    it's a directory). this also protects the strcmp calls below.
		if(!ext)
			continue;
		// .. compiled XML files the engine writes out by the hundreds;
		//    skipping them is a big performance gain.
		if(!stricmp(ext, ".xmb"))
			continue;
		// .. temp files, usually created when an editor saves a file
		//    (delete, create temp, rename temp); no need to reload those.
		if(!stricmp(ext, ".tmp"))
			continue;
		// .. more than one notification for a file; only reload once.
		//    note: this doesn't suffer from the 'reloaded too early'
		//    problem described above; if there's more than one
		//    request in the array, the file has since been written.
		for(uint i = 0; i < num_pending; i++)
			if(!strcmp(pending_reloads[i], V_path))
				continue;

		// path has already been written to pending_reloads,
		// so just mark it valid.
		num_pending++;		
	}

	// rebuild VFS, in case a file that has been changed is currently
	// mounted from an archive (reloading would just grab the unchanged
	// version in the archive). the rebuild sees differing mtimes and
	// always choses the loose file version. only do this once
	// (instead of per reload request) because it's slow (> 1s)!
	if(num_pending != 0)
		CHECK_ERR(rebuild());

	// now actually reload all files in the array we built
	for(uint i = 0; i < num_pending; i++)
		CHECK_ERR(reload_without_rebuild(pending_reloads[i]));

	return 0;
}


//-----------------------------------------------------------------------------






// rationale for not using virtual functions for file_open vs zip_open:
// it would spread out the implementation of each function and makes
// keeping them in sync harder. we will very rarely add new sources and
// all these functions are in one spot anyway.


// given a Mount, return the actual location (portable path) of
// <V_path>. used by vfs_realpath and VFile_reopen.
int x_realpath(const Mount* m, const char* V_exact_path, char* P_real_path)
{
	const char* P_parent_path = 0;

	switch(m->type)
	{
	case MT_ARCHIVE:
		P_parent_path = h_filename(m->archive);
		break;
	case MT_FILE:
		P_parent_path = m->P_name.c_str();
		break;
	default:
		debug_warn(__func__": invalid type");
		return -1;
	}

	const char* remove = m->V_mount_point.c_str();
	const char* replace = P_parent_path;
	return path_replace(P_real_path, V_exact_path, remove, replace);
}



int x_open(const Mount* m, const char* V_exact_path, int flags, TFile* tf, XFile* xf)
{
	// declare variables used in the switch below to avoid needing {}.
	char P_path[PATH_MAX];

	switch(m->type)
	{
	case MT_ARCHIVE:
		if(flags & FILE_WRITE)
		{
			debug_warn(__func__": requesting write access to file in archive");
			return -1;
		}
		RETURN_ERR(zip_open(m->archive, V_exact_path, flags, &xf->u.zf));
		break;
	case MT_FILE:
		CHECK_ERR(x_realpath(m, V_exact_path, P_path));
		RETURN_ERR(file_open(P_path, flags, &xf->u.f));
		break;
	default:
		debug_warn(__func__": invalid type");
		return ERR_CORRUPTED;
	}

	// success
	// note: don't assign these unless we succeed to avoid the
	// false impression that all is well.
	xf->type = m->type;
	xf->tf   = tf;
	return 0;
}


int x_close(XFile* xf)
{
	switch(xf->type)
	{
	// no file open (e.g. because x_open failed) -> nothing to do.
	case MT_NONE:
		return 0;

	case MT_ARCHIVE:
		(void)zip_close(&xf->u.zf);
		break;
	case MT_FILE:
		(void)file_close(&xf->u.f);
		break;
	default:
		debug_warn(__func__": invalid type");
		break;
	}

	// update file state in VFS tree
	// (must be done after close, since that calculates the size)
	if(xf->u.f.flags & FILE_WRITE)	// xxx what about other types?
		tree_update_file(xf->tf, xf->u.f.size, time(0));	// can't fail

	xf->type = MT_NONE;
	return 0;
}


int x_validate(const XFile* xf)
{
	switch(xf->type)
	{
	case MT_NONE:
		if(xf->tf != 0)
			return -100;
		return 0;	// ok, nothing else to check

	case MT_FILE:
		if(xf->tf == 0)
			return -101;
		return file_validate(&xf->u.f);

	case MT_ARCHIVE:
		// this could be set to 0 in x_open (since it's used to update the
		// VFS after newly written files are closed, but archive files
		// cannot be modified), but it's not ATM.
		if(xf->tf == 0)
			return -102;
		return zip_validate(&xf->u.zf);

	default:
		return -103;	// invalid type
	}
	UNREACHABLE;
}


bool x_is_open(const XFile* xf)
{
	return (xf->type != MT_NONE);
}


cassert(offsetof(struct File, size ) == offsetof(struct ZFile, ucsize));
cassert(offsetof(struct File, flags) == offsetof(struct ZFile, flags));

// VFile was exceeding HDATA_USER_SIZE. flags and size (required
// in File as well as VFile) are now moved into the union.
// use the functions below to insulate against change a bit.

off_t x_size(const XFile* xf)
{
	return xf->u.f.size;
}


void x_set_flags(XFile* xf, uint flags)
{
	xf->u.f.flags = flags;
}

uint x_flags(const XFile* xf)
{
	return xf->u.f.flags;
}



int x_io(XFile* xf, off_t ofs, size_t size, void* buf, FileIOCB cb, uintptr_t ctx)
{
	switch(xf->type)
	{
	case MT_ARCHIVE:
		// (vfs_open makes sure it's not opened for writing if zip)
		return zip_read(&xf->u.zf, ofs, size, buf, cb, ctx);

	case MT_FILE:
		// normal file:
		// let file_io alloc the buffer if the caller didn't (i.e. p = 0),
		// because it knows about alignment / padding requirements
		return file_io(&xf->u.f, ofs, size, buf, cb, ctx);

	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_map(XFile* xf, void*& p, size_t& size)
{
	switch(xf->type)
	{
	case MT_ARCHIVE:
		return zip_map(&xf->u.zf, p, size);
	case MT_FILE:
		return file_map(&xf->u.f, p, size);
	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_unmap(XFile* xf)
{
	switch(xf->type)
	{
	case MT_ARCHIVE:
		return zip_unmap(&xf->u.zf);
	case MT_FILE:
		return file_unmap(&xf->u.f);
	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_io_issue(XFile* xf, off_t ofs, size_t size, void* buf, XIo* xio)
{
	xio->type = xf->type;
	switch(xio->type)
	{
	case MT_ARCHIVE:
		return zip_io_issue(&xf->u.zf, ofs, size, buf, &xio->u.zio);
	case MT_FILE:
		return file_io_issue(&xf->u.f, ofs, size, buf, &xio->u.fio);
	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_io_has_completed(XIo* xio)
{
	switch(xio->type)
	{
	case MT_ARCHIVE:
		return zip_io_has_completed(&xio->u.zio);
	case MT_FILE:
		return file_io_has_completed(&xio->u.fio);
	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_io_wait(XIo* xio, void*& p, size_t& size)
{
	switch(xio->type)
	{
	case MT_ARCHIVE:
		return zip_io_wait(&xio->u.zio, p, size);
	case MT_FILE:
		return file_io_wait(&xio->u.fio, p, size);
	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_io_discard(XIo* xio)
{
	switch(xio->type)
	{
	case MT_ARCHIVE:
		return zip_io_discard(&xio->u.zio);
	case MT_FILE:
		return file_io_discard(&xio->u.fio);
	default:
		debug_warn(__func__": invalid file type");
		return ERR_CORRUPTED;
	}
}


int x_io_validate(const XIo* xio)
{
	switch(xio->type)
	{
	case MT_ARCHIVE:
		return zip_io_validate(&xio->u.zio);
	case MT_FILE:
		return file_io_validate(&xio->u.fio);
	default:
		return -100;	// invalid type
	}
	UNREACHABLE;
}