fix locking issue with new mempool limiting #6889

pull jonasschnelli wants to merge 1 commits into bitcoin:master from jonasschnelli:2015/10/fix_mempool_lock changing 2 files +9 −3

jonasschnelli commented at 1:57 PM on October 26, 2015: contributor

Current master crashes on OSX with an exception: "boost: mutex lock failed in pthread_mutex_lock: Invalid argument".

mempool is a global object and gets initialized over cxx_global_var_init(). Calling LOCK() within constructor (of a object in global scope) is problematic (at least on OSX).

fix locking issue with new mempool limiting

Current master crashes on OSX with an exception: "boost: mutex lock failed in pthread_mutex_lock: Invalid argument"

0d699fc821

jonasschnelli force-pushed on Oct 26, 2015
laanwj added the label Bug on Oct 26, 2015
laanwj commented at 4:36 PM on October 26, 2015: member

utACK
jonasschnelli commented at 8:25 PM on October 26, 2015: contributor

The issue that this PR solves was introduced in #6722
petertodd commented at 9:24 PM on October 26, 2015: contributor

utACK
in src/txmempool.h:None in 0d699fc821
```
 374 | @@ -375,6 +375,7 @@ class CTxMemPool
 375 |      void removeForBlock(const std::vector<CTransaction>& vtx, unsigned int nBlockHeight,
 376 |                          std::list<CTransaction>& conflicts, bool fCurrentEstimate = true);
 377 |      void clear();
```
dcousens commented at 1:20 AM on October 27, 2015:

nit: lockAndClear/clearSync maybe?

laanwj commented at 7:41 AM on October 27, 2015:

Usually projects have some convention for this (eg _XX and XX, or XX_nolock and XX) . Where one is used internally (when lock already held) and the other externally (locks and calls inner function). Preferable to have the names as similar as possible with just the lock/nonlock difference. lockAndClear and clearSync are too far apart to be easily recognizable as a set. So IMO the current naming is better.

dcousens commented at 10:28 PM on October 27, 2015:

:+1: just not something I'd seen before.
dcousens commented at 1:20 AM on October 27, 2015: contributor

utACK
laanwj merged this on Oct 27, 2015
laanwj closed this on Oct 27, 2015
laanwj referenced this in commit 38369dda32 on Oct 27, 2015
jnewbery cross-referenced this on Nov 13, 2019 from issue Turn the global mempool into a shared pointer by jnewbery
random-zebra cross-referenced this on May 26, 2020 from issue [Core] MemPool package tracking and limits by random-zebra
furszy referenced this in commit eb00d0f62f on Jun 14, 2020
str4d cross-referenced this on Apr 9, 2021 from issue Mutex lock failure results in crash on exit on native macOS or in debug mode by rex4539
str4d cross-referenced this on Aug 4, 2021 from issue ZIP 239 preparations 1 by str4d
zkbot referenced this in commit 65dff184f6 on Aug 4, 2021
zkbot referenced this in commit 6f38962577 on Aug 5, 2021
zkbot referenced this in commit 1aa59fd4ac on Aug 5, 2021
zkbot referenced this in commit 8c3d670c8a on Aug 9, 2021
zkbot referenced this in commit 1dc1ababa9 on Aug 9, 2021
zkbot referenced this in commit 18fe0a377b on Aug 9, 2021
zkbot referenced this in commit 1cb1ed2653 on Aug 10, 2021
bitcoin locked this on Sep 8, 2021

Contributors

Labels

Linked (view graph)

#6722 Limit mempool by throwing away the cheapest txn and setting min relay fee to it