view src/lib-index/mail-index-sync.c @ 9191:b340ecb24469 HEAD

Fix VPATH build of RQUOTA support. Some rpcgen derive #include "..." paths from the infile argument. This will be off for VPATH builds, as the generated rquota_xdr.c code will look in $(srcdir), but we'll generate the rquota.h file in $(builddir). Play safe and copy rquota.x to $(builddir) first. This fixes the build on openSUSE 11.1.
author Matthias Andree <matthias.andree@gmx.de>
date Tue, 07 Jul 2009 21:01:36 +0200
parents 6cc8ee6080ed
children 9445831aebc0
line wrap: on
line source

/* Copyright (c) 2003-2009 Dovecot authors, see the included COPYING file */

#include "lib.h"
#include "array.h"
#include "mail-index-view-private.h"
#include "mail-index-sync-private.h"
#include "mail-index-transaction-private.h"
#include "mail-transaction-log-private.h"
#include "mail-cache.h"

#include <stdio.h>
#include <stdlib.h>

struct mail_index_sync_ctx {
	struct mail_index *index;
	struct mail_index_view *view;
	struct mail_index_transaction *sync_trans, *ext_trans;
	enum mail_index_sync_flags flags;

	const struct mail_transaction_header *hdr;
	const void *data;

	ARRAY_DEFINE(sync_list, struct mail_index_sync_list);
	uint32_t next_uid;
	uint32_t last_tail_seq, last_tail_offset;

	uint32_t append_uid_first, append_uid_last;

	unsigned int sync_appends:1;
};

static void mail_index_sync_add_expunge(struct mail_index_sync_ctx *ctx)
{
	const struct mail_transaction_expunge *e = ctx->data;
	size_t i, size = ctx->hdr->size / sizeof(*e);
	uint32_t uid;

	for (i = 0; i < size; i++) {
		for (uid = e[i].uid1; uid <= e[i].uid2; uid++)
			mail_index_expunge(ctx->sync_trans, uid);
	}
}

static void mail_index_sync_add_flag_update(struct mail_index_sync_ctx *ctx)
{
	const struct mail_transaction_flag_update *u = ctx->data;
	size_t i, size = ctx->hdr->size / sizeof(*u);

	for (i = 0; i < size; i++) {
		if (u[i].add_flags != 0) {
			mail_index_update_flags_range(ctx->sync_trans,
						      u[i].uid1, u[i].uid2,
						      MODIFY_ADD,
						      u[i].add_flags);
		}
		if (u[i].remove_flags != 0) {
			mail_index_update_flags_range(ctx->sync_trans,
						      u[i].uid1, u[i].uid2,
						      MODIFY_REMOVE,
						      u[i].remove_flags);
		}
	}
}

static void mail_index_sync_add_keyword_update(struct mail_index_sync_ctx *ctx)
{
	const struct mail_transaction_keyword_update *u = ctx->data;
	const char *keyword_names[2];
	struct mail_keywords *keywords;
	const uint32_t *uids;
	uint32_t uid;
	size_t uidset_offset, i, size;

	uidset_offset = sizeof(*u) + u->name_size;
	if ((uidset_offset % 4) != 0)
		uidset_offset += 4 - (uidset_offset % 4);
	uids = CONST_PTR_OFFSET(u, uidset_offset);

	keyword_names[0] = t_strndup(u + 1, u->name_size);
	keyword_names[1] = NULL;
	keywords = mail_index_keywords_create(ctx->index, keyword_names);

	size = (ctx->hdr->size - uidset_offset) / sizeof(uint32_t);
	for (i = 0; i < size; i += 2) {
		/* FIXME: mail_index_update_keywords_range() */
		for (uid = uids[i]; uid <= uids[i+1]; uid++) {
			mail_index_update_keywords(ctx->sync_trans, uid,
						   u->modify_type, keywords);
		}
	}

	mail_index_keywords_free(&keywords);
}

static void mail_index_sync_add_keyword_reset(struct mail_index_sync_ctx *ctx)
{
	const struct mail_transaction_keyword_reset *u = ctx->data;
	size_t i, size = ctx->hdr->size / sizeof(*u);
	struct mail_keywords *keywords;
	uint32_t uid;

	keywords = mail_index_keywords_create(ctx->index, NULL);
	for (i = 0; i < size; i++) {
		for (uid = u[i].uid1; uid <= u[i].uid2; uid++) {
			mail_index_update_keywords(ctx->sync_trans, uid,
						   MODIFY_REPLACE, keywords);
		}
	}
	mail_index_keywords_free(&keywords);
}

static void mail_index_sync_add_append(struct mail_index_sync_ctx *ctx)
{
	const struct mail_index_record *rec = ctx->data;

	if (ctx->append_uid_first == 0 || rec->uid < ctx->append_uid_first)
		ctx->append_uid_first = rec->uid;

	rec = CONST_PTR_OFFSET(ctx->data, ctx->hdr->size - sizeof(*rec));
	if (rec->uid > ctx->append_uid_last)
		ctx->append_uid_last = rec->uid;

	ctx->sync_appends = TRUE;
}

static bool mail_index_sync_add_transaction(struct mail_index_sync_ctx *ctx)
{
	switch (ctx->hdr->type & MAIL_TRANSACTION_TYPE_MASK) {
	case MAIL_TRANSACTION_EXPUNGE:
		mail_index_sync_add_expunge(ctx);
		break;
	case MAIL_TRANSACTION_FLAG_UPDATE:
                mail_index_sync_add_flag_update(ctx);
		break;
	case MAIL_TRANSACTION_KEYWORD_UPDATE:
                mail_index_sync_add_keyword_update(ctx);
		break;
	case MAIL_TRANSACTION_KEYWORD_RESET:
                mail_index_sync_add_keyword_reset(ctx);
		break;
	case MAIL_TRANSACTION_APPEND:
		mail_index_sync_add_append(ctx);
		break;
	default:
		return FALSE;
	}
	return TRUE;
}

static void mail_index_sync_add_dirty_updates(struct mail_index_sync_ctx *ctx)
{
	struct mail_transaction_flag_update update;
	const struct mail_index_record *rec;
	uint32_t seq, messages_count;

	memset(&update, 0, sizeof(update));

	messages_count = mail_index_view_get_messages_count(ctx->view);
	for (seq = 1; seq <= messages_count; seq++) {
		rec = mail_index_lookup(ctx->view, seq);
		if ((rec->flags & MAIL_INDEX_MAIL_FLAG_DIRTY) == 0)
			continue;

		mail_index_update_flags(ctx->sync_trans, rec->uid,
					MODIFY_REPLACE, rec->flags);
	}
}

static void
mail_index_sync_update_mailbox_pos(struct mail_index_sync_ctx *ctx)
{
	uint32_t seq;
	uoff_t offset;

	mail_transaction_log_view_get_prev_pos(ctx->view->log_view,
					       &seq, &offset);

	ctx->last_tail_seq = seq;
	ctx->last_tail_offset = offset + ctx->hdr->size + sizeof(*ctx->hdr);
}

static int
mail_index_sync_read_and_sort(struct mail_index_sync_ctx *ctx)
{
	struct mail_index_transaction *sync_trans = ctx->sync_trans;
	struct mail_index_sync_list *synclist;
        const struct mail_index_transaction_keyword_update *keyword_updates;
	unsigned int i, keyword_count;
	int ret;

	if ((ctx->view->map->hdr.flags & MAIL_INDEX_HDR_FLAG_HAVE_DIRTY) &&
	    (ctx->flags & MAIL_INDEX_SYNC_FLAG_FLUSH_DIRTY) != 0) {
		/* show dirty flags as flag updates */
		mail_index_sync_add_dirty_updates(ctx);
	}

	/* read all transactions from log into a transaction in memory.
	   skip the external ones, they're already synced to mailbox and
	   included in our view */
	while ((ret = mail_transaction_log_view_next(ctx->view->log_view,
						     &ctx->hdr,
						     &ctx->data)) > 0) {
		if ((ctx->hdr->type & MAIL_TRANSACTION_EXTERNAL) != 0)
			continue;

		T_BEGIN {
			if (mail_index_sync_add_transaction(ctx))
				mail_index_sync_update_mailbox_pos(ctx);
		} T_END;
	}

	/* create an array containing all expunge, flag and keyword update
	   arrays so we can easily go through all of the changes. */
	keyword_count = !array_is_created(&sync_trans->keyword_updates) ? 0 :
		array_count(&sync_trans->keyword_updates);
	i_array_init(&ctx->sync_list, keyword_count + 2);

	if (array_is_created(&sync_trans->expunges)) {
		synclist = array_append_space(&ctx->sync_list);
		synclist->array = (void *)&sync_trans->expunges;
	}

	if (array_is_created(&sync_trans->updates)) {
		synclist = array_append_space(&ctx->sync_list);
		synclist->array = (void *)&sync_trans->updates;
	}

	/* we must return resets before keyword additions or they get lost */
	if (array_is_created(&sync_trans->keyword_resets)) {
		synclist = array_append_space(&ctx->sync_list);
		synclist->array = (void *)&sync_trans->keyword_resets;
	}

	keyword_updates = keyword_count == 0 ? NULL :
		array_idx(&sync_trans->keyword_updates, 0);
	for (i = 0; i < keyword_count; i++) {
		if (array_is_created(&keyword_updates[i].add_seq)) {
			synclist = array_append_space(&ctx->sync_list);
			synclist->array = (void *)&keyword_updates[i].add_seq;
			synclist->keyword_idx = i;
		}
		if (array_is_created(&keyword_updates[i].remove_seq)) {
			synclist = array_append_space(&ctx->sync_list);
			synclist->array =
				(void *)&keyword_updates[i].remove_seq;
			synclist->keyword_idx = i;
			synclist->keyword_remove = TRUE;
		}
	}

	return ret;
}

static bool
mail_index_need_sync(struct mail_index *index, enum mail_index_sync_flags flags,
		     uint32_t log_file_seq, uoff_t log_file_offset)
{
	const struct mail_index_header *hdr = &index->map->hdr;
	if ((flags & MAIL_INDEX_SYNC_FLAG_REQUIRE_CHANGES) == 0)
		return TRUE;

	/* sync only if there's something to do */
	if (hdr->first_recent_uid < hdr->next_uid &&
	    (flags & MAIL_INDEX_SYNC_FLAG_DROP_RECENT) != 0)
		return TRUE;

	if ((hdr->flags & MAIL_INDEX_HDR_FLAG_HAVE_DIRTY) &&
	    (flags & MAIL_INDEX_SYNC_FLAG_FLUSH_DIRTY) != 0)
		return TRUE;

	if (log_file_seq == (uint32_t)-1) {
		/* we want to sync up to transaction log's head */
		mail_transaction_log_get_head(index->log,
					      &log_file_seq, &log_file_offset);
	}
	if ((hdr->log_file_tail_offset < log_file_offset &&
	     hdr->log_file_seq == log_file_seq) ||
	    hdr->log_file_seq < log_file_seq)
		return TRUE;

	if (index->need_recreate)
		return TRUE;

	/* already synced */
	return mail_cache_need_compress(index->cache);
}

static int
mail_index_sync_set_log_view(struct mail_index_view *view,
			     uint32_t start_file_seq, uoff_t start_file_offset)
{
	uint32_t log_seq;
	uoff_t log_offset;
	bool reset;
	int ret;

	mail_transaction_log_get_head(view->index->log, &log_seq, &log_offset);

	ret = mail_transaction_log_view_set(view->log_view,
                                            start_file_seq, start_file_offset,
					    log_seq, log_offset, &reset);
	if (ret <= 0) {
		/* either corrupted or the file was deleted for
		   some reason. either way, we can't go forward */
		mail_index_set_error(view->index,
			"Unexpected transaction log desync with index %s",
			view->index->filepath);
		return -1;
	}
	return 0;
}

int mail_index_sync_begin(struct mail_index *index,
			  struct mail_index_sync_ctx **ctx_r,
			  struct mail_index_view **view_r,
			  struct mail_index_transaction **trans_r,
			  enum mail_index_sync_flags flags)
{
	int ret;

	ret = mail_index_sync_begin_to(index, ctx_r, view_r, trans_r,
				       (uint32_t)-1, (uoff_t)-1, flags);
	i_assert(ret != 0 ||
		 (flags & MAIL_INDEX_SYNC_FLAG_REQUIRE_CHANGES) != 0);
	return ret;
}

static int
mail_index_sync_begin_init(struct mail_index *index,
			   enum mail_index_sync_flags flags,
			   uint32_t log_file_seq, uoff_t log_file_offset)
{
	const struct mail_index_header *hdr;
	uint32_t seq;
	uoff_t offset;
	bool locked = FALSE;
	int ret;

	/* if we require changes, don't lock transaction log yet. first check
	   if there's anything to sync. */
	if ((flags & MAIL_INDEX_SYNC_FLAG_REQUIRE_CHANGES) == 0) {
		if (mail_transaction_log_sync_lock(index->log,
						   &seq, &offset) < 0)
			return -1;
		locked = TRUE;
	}

	/* The view must contain what we expect the mailbox to look like
	   currently. That allows the backend to update external flag
	   changes (etc.) if the view doesn't match the mailbox.

	   We'll update the view to contain everything that exist in the
	   transaction log except for expunges. They're synced in
	   mail_index_sync_commit(). */
	if ((ret = mail_index_map(index, MAIL_INDEX_SYNC_HANDLER_HEAD)) <= 0) {
		if (ret == 0) {
			if (locked)
				mail_transaction_log_sync_unlock(index->log);
			return -1;
		}

		/* let's try again */
		if (mail_index_map(index, MAIL_INDEX_SYNC_HANDLER_HEAD) <= 0) {
			if (locked)
				mail_transaction_log_sync_unlock(index->log);
			return -1;
		}
	}

	if (!mail_index_need_sync(index, flags,
				  log_file_seq, log_file_offset)) {
		if (locked)
			mail_transaction_log_sync_unlock(index->log);
		return 0;
	}

	if (!locked) {
		/* it looks like we have something to sync. lock the file and
		   check again. */
		flags &= ~MAIL_INDEX_SYNC_FLAG_REQUIRE_CHANGES;
		return mail_index_sync_begin_init(index, flags, log_file_seq,
						  log_file_offset);
	}

	hdr = &index->map->hdr;
	if (hdr->log_file_tail_offset > hdr->log_file_head_offset ||
	    hdr->log_file_seq > seq ||
	    (hdr->log_file_seq == seq && hdr->log_file_tail_offset > offset)) {
		/* broken sync positions. fix them. */
		mail_index_set_error(index,
			"broken sync positions in index file %s",
			index->filepath);
		mail_index_fsck_locked(index);
	}
	return 1;
}

int mail_index_sync_begin_to(struct mail_index *index,
			     struct mail_index_sync_ctx **ctx_r,
			     struct mail_index_view **view_r,
			     struct mail_index_transaction **trans_r,
			     uint32_t log_file_seq, uoff_t log_file_offset,
			     enum mail_index_sync_flags flags)
{
	const struct mail_index_header *hdr;
	struct mail_index_sync_ctx *ctx;
	struct mail_index_view *sync_view;
	enum mail_index_transaction_flags trans_flags;
	int ret;

	i_assert(!index->syncing);

	if (log_file_seq != (uint32_t)-1)
		flags |= MAIL_INDEX_SYNC_FLAG_REQUIRE_CHANGES;

	ret = mail_index_sync_begin_init(index, flags, log_file_seq,
					 log_file_offset);
	if (ret <= 0)
		return ret;

	hdr = &index->map->hdr;

	ctx = i_new(struct mail_index_sync_ctx, 1);
	ctx->index = index;
	ctx->last_tail_seq = hdr->log_file_seq;
	ctx->last_tail_offset = hdr->log_file_tail_offset;
	ctx->flags = flags;

	ctx->view = mail_index_view_open(index);

	sync_view = mail_index_dummy_view_open(index);
	ctx->sync_trans = mail_index_transaction_begin(sync_view,
					MAIL_INDEX_TRANSACTION_FLAG_EXTERNAL);
	mail_index_view_close(&sync_view);

	/* we wish to see all the changes from last mailbox sync position to
	   the end of the transaction log */
	if (mail_index_sync_set_log_view(ctx->view, hdr->log_file_seq,
					 hdr->log_file_tail_offset) < 0) {
		/* if a log file is missing, there's nothing we can do except
		   to skip over it. fix the problem with fsck and try again. */
		mail_index_fsck_locked(index);
		mail_index_sync_rollback(&ctx);
		return mail_index_sync_begin_to(index, ctx_r, view_r, trans_r,
						log_file_seq, log_file_offset,
						flags);
	}

	/* we need to have all the transactions sorted to optimize
	   caller's mailbox access patterns */
	if (mail_index_sync_read_and_sort(ctx) < 0) {
                mail_index_sync_rollback(&ctx);
		return -1;
	}

	ctx->view->index_sync_view = TRUE;

	/* create the transaction after the view has been updated with
	   external transactions and marked as sync view */
	trans_flags = MAIL_INDEX_TRANSACTION_FLAG_EXTERNAL;
	if ((ctx->flags & MAIL_INDEX_SYNC_FLAG_AVOID_FLAG_UPDATES) != 0)
		trans_flags |= MAIL_INDEX_TRANSACTION_FLAG_AVOID_FLAG_UPDATES;
	ctx->ext_trans = mail_index_transaction_begin(ctx->view, trans_flags);
	ctx->ext_trans->sync_transaction = TRUE;

	index->syncing = TRUE;

	*ctx_r = ctx;
	*view_r = ctx->view;
	*trans_r = ctx->ext_trans;
	return 1;
}

static bool mail_index_sync_view_have_any(struct mail_index_view *view,
					  enum mail_index_sync_flags flags)
{
	const struct mail_transaction_header *hdr;
	const void *data;
	uint32_t log_seq;
	uoff_t log_offset;
	bool reset;
	int ret;

	if (view->map->hdr.first_recent_uid < view->map->hdr.next_uid &&
	    (flags & MAIL_INDEX_SYNC_FLAG_DROP_RECENT) != 0)
		return TRUE;

	if ((view->map->hdr.flags & MAIL_INDEX_HDR_FLAG_HAVE_DIRTY) &&
	    (flags & MAIL_INDEX_SYNC_FLAG_FLUSH_DIRTY) != 0)
		return TRUE;

	mail_transaction_log_get_head(view->index->log, &log_seq, &log_offset);
	if (mail_transaction_log_view_set(view->log_view,
					  view->map->hdr.log_file_seq,
					  view->map->hdr.log_file_tail_offset,
					  log_seq, log_offset, &reset) <= 0) {
		/* let the actual syncing handle the error */
		return TRUE;
	}

	while ((ret = mail_transaction_log_view_next(view->log_view,
						     &hdr, &data)) > 0) {
		if ((hdr->type & MAIL_TRANSACTION_EXTERNAL) != 0)
			continue;

		switch (hdr->type & MAIL_TRANSACTION_TYPE_MASK) {
		case MAIL_TRANSACTION_EXT_REC_UPDATE:
			/* extension record updates aren't exactly needed
			   to be synced, but cache syncing relies on tail
			   offsets being updated. */
		case MAIL_TRANSACTION_EXPUNGE:
		case MAIL_TRANSACTION_FLAG_UPDATE:
		case MAIL_TRANSACTION_KEYWORD_UPDATE:
		case MAIL_TRANSACTION_KEYWORD_RESET:
			return TRUE;
		default:
			break;
		}
	}
	return ret < 0;
}

bool mail_index_sync_have_any(struct mail_index *index,
			      enum mail_index_sync_flags flags)
{
	struct mail_index_view *view;
	bool ret;

	(void)mail_index_refresh(index);

	view = mail_index_view_open(index);
	ret = mail_index_sync_view_have_any(view, flags);
	mail_index_view_close(&view);
	return ret;
}

static void
mail_index_sync_get_expunge(struct mail_index_sync_rec *rec,
			    const struct mail_transaction_expunge *exp)
{
	rec->type = MAIL_INDEX_SYNC_TYPE_EXPUNGE;
	rec->uid1 = exp->uid1;
	rec->uid2 = exp->uid2;
}

static void
mail_index_sync_get_update(struct mail_index_sync_rec *rec,
			   const struct mail_transaction_flag_update *update)
{
	rec->type = MAIL_INDEX_SYNC_TYPE_FLAGS;
	rec->uid1 = update->uid1;
	rec->uid2 = update->uid2;

	rec->add_flags = update->add_flags;
	rec->remove_flags = update->remove_flags;
}

static void
mail_index_sync_get_keyword_update(struct mail_index_sync_rec *rec,
				   const struct uid_range *range,
				   struct mail_index_sync_list *sync_list)
{
	rec->type = !sync_list->keyword_remove ?
		MAIL_INDEX_SYNC_TYPE_KEYWORD_ADD :
		MAIL_INDEX_SYNC_TYPE_KEYWORD_REMOVE;
	rec->uid1 = range->uid1;
	rec->uid2 = range->uid2;
	rec->keyword_idx = sync_list->keyword_idx;
}

static void mail_index_sync_get_keyword_reset(struct mail_index_sync_rec *rec,
					       const struct uid_range *range)
{
	rec->type = MAIL_INDEX_SYNC_TYPE_KEYWORD_RESET;
	rec->uid1 = range->uid1;
	rec->uid2 = range->uid2;
}

bool mail_index_sync_next(struct mail_index_sync_ctx *ctx,
			  struct mail_index_sync_rec *sync_rec)
{
	struct mail_index_transaction *sync_trans = ctx->sync_trans;
	struct mail_index_sync_list *sync_list;
	const struct uid_range *uid_range = NULL;
	unsigned int i, count, next_i;
	uint32_t next_found_uid;

	next_i = (unsigned int)-1;
	next_found_uid = (uint32_t)-1;

	/* FIXME: replace with a priority queue so we don't have to go
	   through the whole list constantly. and remember to make sure that
	   keyword resets are sent before adds! */
	sync_list = array_get_modifiable(&ctx->sync_list, &count);
	for (i = 0; i < count; i++) {
		if (!array_is_created(sync_list[i].array) ||
		    sync_list[i].idx == array_count(sync_list[i].array))
			continue;

		uid_range = array_idx(sync_list[i].array, sync_list[i].idx);
		if (uid_range->uid1 == ctx->next_uid) {
			/* use this one. */
			break;
		}
		if (uid_range->uid1 < next_found_uid) {
			next_i = i;
                        next_found_uid = uid_range->uid1;
		}
	}

	if (i == count) {
		if (next_i == (unsigned int)-1) {
			/* nothing left in sync_list */
			if (ctx->sync_appends) {
				ctx->sync_appends = FALSE;
				sync_rec->type = MAIL_INDEX_SYNC_TYPE_APPEND;
				sync_rec->uid1 = ctx->append_uid_first;
				sync_rec->uid2 = ctx->append_uid_last;
				return TRUE;
			}
			return FALSE;
		}
                ctx->next_uid = next_found_uid;
		i = next_i;
		uid_range = array_idx(sync_list[i].array, sync_list[i].idx);
	}

	if (sync_list[i].array == (void *)&sync_trans->expunges) {
		mail_index_sync_get_expunge(sync_rec,
			(const struct mail_transaction_expunge *)uid_range);
	} else if (sync_list[i].array == (void *)&sync_trans->updates) {
		mail_index_sync_get_update(sync_rec,
			(const struct mail_transaction_flag_update *)uid_range);
	} else if (sync_list[i].array == (void *)&sync_trans->keyword_resets) {
		mail_index_sync_get_keyword_reset(sync_rec, uid_range);
	} else {
		mail_index_sync_get_keyword_update(sync_rec, uid_range,
						   &sync_list[i]);
	}
	sync_list[i].idx++;
	return TRUE;
}

bool mail_index_sync_have_more(struct mail_index_sync_ctx *ctx)
{
	const struct mail_index_sync_list *sync_list;
	unsigned int i, count;

	if (ctx->sync_appends)
		return TRUE;

	sync_list = array_get(&ctx->sync_list, &count);
	for (i = 0; i < count; i++) {
		if (array_is_created(sync_list[i].array) &&
		    sync_list[i].idx != array_count(sync_list[i].array))
			return TRUE;
	}
	return FALSE;
}

void mail_index_sync_reset(struct mail_index_sync_ctx *ctx)
{
	struct mail_index_sync_list *sync_list;
	unsigned int i, count;

	ctx->next_uid = 0;

	sync_list = array_get_modifiable(&ctx->sync_list, &count);
	for (i = 0; i < count; i++)
		sync_list[i].idx = 0;
}

static void mail_index_sync_end(struct mail_index_sync_ctx **_ctx)
{
        struct mail_index_sync_ctx *ctx = *_ctx;

	i_assert(ctx->index->syncing);

	*_ctx = NULL;

	ctx->index->syncing = FALSE;
	mail_transaction_log_sync_unlock(ctx->index->log);

	mail_index_view_close(&ctx->view);
	mail_index_transaction_rollback(&ctx->sync_trans);
	if (array_is_created(&ctx->sync_list))
		array_free(&ctx->sync_list);
	i_free(ctx);
}

static void
mail_index_sync_update_mailbox_offset(struct mail_index_sync_ctx *ctx)
{
	uint32_t seq;
	uoff_t offset;

	mail_transaction_log_view_get_prev_pos(ctx->view->log_view,
					       &seq, &offset);
	mail_transaction_log_set_mailbox_sync_pos(ctx->index->log, seq, offset);

	/* If tail offset has changed, make sure it gets written to
	   transaction log. */
	if (ctx->last_tail_offset != offset)
		ctx->ext_trans->log_updates = TRUE;
}

static bool mail_index_sync_want_index_write(struct mail_index *index)
{
	uint32_t log_diff;

	log_diff = index->map->hdr.log_file_tail_offset -
		index->last_read_log_file_tail_offset;
	if (log_diff > MAIL_INDEX_MAX_WRITE_BYTES ||
	    (index->index_min_write && log_diff > MAIL_INDEX_MIN_WRITE_BYTES))
		return TRUE;

	if (index->need_recreate)
		return TRUE;
	return FALSE;
}

int mail_index_sync_commit(struct mail_index_sync_ctx **_ctx)
{
        struct mail_index_sync_ctx *ctx = *_ctx;
	struct mail_index *index = ctx->index;
	uint32_t seq, next_uid;
	uoff_t offset;
	bool want_rotate;
	int ret = 0;

	mail_index_sync_update_mailbox_offset(ctx);
	if (mail_cache_need_compress(index->cache)) {
		/* if cache compression fails, we don't really care.
		   the cache offsets are updated only if the compression was
		   successful. */
		(void)mail_cache_compress(index->cache, ctx->ext_trans);
	}

	if ((ctx->flags & MAIL_INDEX_SYNC_FLAG_DROP_RECENT) != 0) {
		next_uid = mail_index_transaction_get_next_uid(ctx->ext_trans);
		if (index->map->hdr.first_recent_uid < next_uid) {
			mail_index_update_header(ctx->ext_trans,
				offsetof(struct mail_index_header,
					 first_recent_uid),
				&next_uid, sizeof(next_uid), FALSE);
		}
	}

	if (mail_index_transaction_commit(&ctx->ext_trans, &seq, &offset) < 0) {
		mail_index_sync_end(&ctx);
		return -1;
	}

	/* refresh the mapping with newly committed external transactions
	   and the synced expunges. sync using file handler here so that the
	   expunge handlers get called. */
	if (mail_index_map(ctx->index, MAIL_INDEX_SYNC_HANDLER_FILE) <= 0)
		ret = -1;

	want_rotate = mail_transaction_log_want_rotate(index->log);
	if (ret == 0 &&
	    (want_rotate || mail_index_sync_want_index_write(index))) {
		index->need_recreate = FALSE;
		index->index_min_write = FALSE;
		mail_index_write(index, want_rotate);
	}
	mail_index_sync_end(_ctx);
	return ret;
}

void mail_index_sync_rollback(struct mail_index_sync_ctx **ctx)
{
	if ((*ctx)->ext_trans != NULL)
		mail_index_transaction_rollback(&(*ctx)->ext_trans);
	mail_index_sync_end(ctx);
}

void mail_index_sync_flags_apply(const struct mail_index_sync_rec *sync_rec,
				 uint8_t *flags)
{
	i_assert(sync_rec->type == MAIL_INDEX_SYNC_TYPE_FLAGS);

	*flags = (*flags & ~sync_rec->remove_flags) | sync_rec->add_flags;
}

bool mail_index_sync_keywords_apply(const struct mail_index_sync_rec *sync_rec,
				    ARRAY_TYPE(keyword_indexes) *keywords)
{
	const unsigned int *keyword_indexes;
	unsigned int idx = sync_rec->keyword_idx;
	unsigned int i, count;

	keyword_indexes = array_get(keywords, &count);
	switch (sync_rec->type) {
	case MAIL_INDEX_SYNC_TYPE_KEYWORD_ADD:
		for (i = 0; i < count; i++) {
			if (keyword_indexes[i] == idx)
				return FALSE;
		}

		array_append(keywords, &idx, 1);
		return TRUE;
	case MAIL_INDEX_SYNC_TYPE_KEYWORD_REMOVE:
		for (i = 0; i < count; i++) {
			if (keyword_indexes[i] == idx) {
				array_delete(keywords, i, 1);
				return TRUE;
			}
		}
		return FALSE;
	case MAIL_INDEX_SYNC_TYPE_KEYWORD_RESET:
		if (array_count(keywords) == 0)
			return FALSE;

		array_clear(keywords);
		return TRUE;
	default:
		i_unreached();
		return FALSE;
	}
}

void mail_index_sync_set_corrupted(struct mail_index_sync_map_ctx *ctx,
				   const char *fmt, ...)
{
	va_list va;
	uint32_t seq;
	uoff_t offset;

	ctx->errors = TRUE;

	mail_transaction_log_view_get_prev_pos(ctx->view->log_view,
					       &seq, &offset);

	if (seq < ctx->view->index->fsck_log_head_file_seq ||
	    (seq == ctx->view->index->fsck_log_head_file_seq &&
	     offset < ctx->view->index->fsck_log_head_file_offset)) {
		/* be silent */
		return;
	}

	va_start(va, fmt);
	T_BEGIN {
		mail_index_set_error(ctx->view->index,
				     "Log synchronization error at "
				     "seq=%u,offset=%"PRIuUOFF_T" for %s: %s",
				     seq, offset, ctx->view->index->filepath,
				     t_strdup_vprintf(fmt, va));
	} T_END;
	va_end(va);
}