From: Aleksandr Lyapunov <alyapunov@tarantool.org> To: Nikita Pettik <korablev@tarantool.org> Cc: tarantool-patches@dev.tarantool.org Subject: Re: [Tarantool-patches] [PATCH v4 12/12] txm: add a test Date: Tue, 22 Sep 2020 20:58:14 +0300 [thread overview] Message-ID: <acc65f1a-37f2-36d8-8361-6da1c2d6f4f5@tarantool.org> (raw) In-Reply-To: <20200915180523.GD23208@tarantool.org> Hi, thanks for the review. I added the tests (triggers and recovery I'll do a bit later). Found some bugs, here the fixes I made to make the tests pass. diff --git a/src/box/memtx_tx.c b/src/box/memtx_tx.c index d72fa92..5aacd22 100644 --- a/src/box/memtx_tx.c +++ b/src/box/memtx_tx.c @@ -422,7 +422,7 @@ memtx_tx_story_gc_step() * actually delete the tuple, it must be deleted from * index. */ - if (story->del_psn > 0) { + if (story->del_psn > 0 && story->space != NULL) { struct index *index = story->space->index[i]; struct tuple *unused; if (index_replace(index, story->tuple, NULL, @@ -801,7 +801,6 @@ memtx_tx_history_rollback_stmt(struct txn_stmt *stmt) memtx_tx_story_unlink(story, i); } stmt->add_story->add_stmt = NULL; - memtx_tx_story_delete(stmt->add_story); stmt->add_story = NULL; tuple_unref(stmt->new_tuple); } @@ -945,8 +944,22 @@ memtx_tx_history_prepare_stmt(struct txn_stmt *stmt) if (stmt->add_story != NULL) stmt->add_story->add_psn = stmt->txn->psn; - if (stmt->del_story != NULL) + if (stmt->del_story != NULL) { stmt->del_story->del_psn = stmt->txn->psn; + // Let's conflict all other deleting stories. + struct txn_stmt *dels = stmt->del_story->del_stmt; + while (dels != NULL) { + struct txn_stmt *next = dels->next_in_del_list; + if (dels != stmt) { + dels->del_story = NULL; + dels->next_in_del_list = NULL; + } + dels = next; + } + // Set the only deleting statement for that story. + stmt->del_story->del_stmt = stmt; + stmt->next_in_del_list = NULL; + } } ssize_t @@ -1001,6 +1014,20 @@ memtx_tx_tuple_clarify_slow(struct txn *txn, struct space *space, return result; } +void +memtx_tx_on_space_delete(struct space *space) +{ + /* Just clear pointer to space, it will be handled in GC. */ + while (!rlist_empty(&space->memtx_stories)) { + struct memtx_story *story + = rlist_first_entry(&space->memtx_stories, + struct memtx_story, + in_space_stories); + story->space = NULL; + rlist_del(&story->in_space_stories); + } +} + static void memtx_tx_story_delete(struct memtx_story *story) { diff --git a/src/box/memtx_tx.h b/src/box/memtx_tx.h index 6197d1b..aa204ac 100644 --- a/src/box/memtx_tx.h +++ b/src/box/memtx_tx.h @@ -310,6 +310,15 @@ memtx_tx_tuple_clarify(struct txn *txn, struct space *space, is_prepared_ok); } +/** + * Notify manager the a space is deleted. + * It's necessary because there is a chance that garbage collector hasn't + * deleted all stories of that space and in that case some actions of + * story's destructor are not applicable. + */ +void +memtx_tx_on_space_delete(struct space *space); + #if defined(__cplusplus) } /* extern "C" */ #endif /* defined(__cplusplus) */ diff --git a/src/box/space.c b/src/box/space.c index 1243932..6d1d771 100644 --- a/src/box/space.c +++ b/src/box/space.c @@ -38,6 +38,7 @@ #include "user.h" #include "session.h" #include "txn.h" +#include "memtx_tx.h" #include "tuple.h" #include "xrow_update.h" #include "request.h" @@ -253,7 +254,7 @@ space_new_ephemeral(struct space_def *def, struct rlist *key_list) void space_delete(struct space *space) { - rlist_del(&space->memtx_stories); + memtx_tx_on_space_delete(space); assert(space->ck_constraint_trigger == NULL); for (uint32_t j = 0; j <= space->index_id_max; j++) { struct index *index = space->index_map[j]; On 15.09.2020 21:05, Nikita Pettik wrote: > Uncovered test scenarious: more than two indexes; more than one space; > unique secondary indexes; recovery; triggers; rollbacks... >
next prev parent reply other threads:[~2020-09-22 17:58 UTC|newest] Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-09-08 10:22 [Tarantool-patches] [PATCH v4 00/12] Transaction engine for memtx engine Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 01/12] vinyl: rename tx_manager -> vy_tx_manager Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 02/12] txm: add TX status Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 03/12] txm: save does_require_old_tuple flag in txn_stmt Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 04/12] txm: introduce prepare sequence number Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 05/12] txm: introduce memtx tx manager Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 06/12] txm: introduce conflict tracker Aleksandr Lyapunov 2020-09-14 16:36 ` Nikita Pettik 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 07/12] txm: introduce memtx_story Aleksandr Lyapunov 2020-09-15 14:33 ` Nikita Pettik 2020-09-22 17:51 ` Aleksandr Lyapunov 2020-09-23 10:25 ` Nikita Pettik 2020-09-23 11:09 ` Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 08/12] txm: introduce snapshot cleaner Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 09/12] txm: clarify all fetched tuples Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 10/12] txm: use new tx manager in memtx Aleksandr Lyapunov 2020-09-15 17:59 ` Nikita Pettik 2020-09-22 17:53 ` Aleksandr Lyapunov 2020-09-23 10:26 ` Nikita Pettik 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 11/12] test: move txn_proxy.lua to box/lua Aleksandr Lyapunov 2020-09-08 10:22 ` [Tarantool-patches] [PATCH v4 12/12] txm: add a test Aleksandr Lyapunov 2020-09-15 18:05 ` Nikita Pettik 2020-09-22 17:58 ` Aleksandr Lyapunov [this message] 2020-09-23 11:07 ` Nikita Pettik 2020-09-23 11:12 ` Aleksandr Lyapunov 2020-09-23 12:18 ` [Tarantool-patches] [PATCH v4 00/12] Transaction engine for memtx engine Kirill Yukhin
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=acc65f1a-37f2-36d8-8361-6da1c2d6f4f5@tarantool.org \ --to=alyapunov@tarantool.org \ --cc=korablev@tarantool.org \ --cc=tarantool-patches@dev.tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH v4 12/12] txm: add a test' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox