fredrikwidlund / libdynamic Goto Github PK

Hey, can you create a tag for version 2.1, even something like 2.1-dev or 2.1-20201214. Would be useful to have a stable way to easily clone it.

We use your code for our system (K3), and so far have gotten great performance boosts out of it. I just tried switching my data structures from C++'s std::unordered_map to mapi's, and even though it worked for other uses, performance was really bad. perf showed that mapi_erase is very slow since it must shift the whole array with every erase. Is there any way to improve this?

Is this library licensed under GPL or LGPL?

Hi Fredrik,
I would like to add your implementation to my current project [rperf] at
https://github.com/rcarbone/rperf
but I am experimenting issues and core dumps due to invalid parameters passed in the equal() callback
while trying to add items to the map.

With 5 items all seems to be ok as in:
[email protected] 10904> ./issue
libdynamic map testsuite for checking memory with valgrind
Objs Alloc/Free ............... Ok
Alloc/Insert/Free ............
0001: Adding ["key-0001" / 0x7fc25879b590] Ok
0002: Adding ["key-0002" / 0x7fc25879b540] Ok
0003: Adding ["key-0003" / 0x7fc25879b4f0] Ok
0004: Adding ["key-0004" / 0x7fc25879b4a0] Ok
0005: Adding ["key-0005" / 0x7fc25879b450] Ok
Ok

and also valgrind reports no error as in:
[email protected] 10905> valgrind ./issue
==23759== Memcheck, a memory error detector
==23759== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==23759== Using Valgrind-3.12.0 and LibVEX; rerun with -h for copyright info
==23759== Command: ./issue
==23759==
libdynamic map testsuite for checking memory with valgrind
Objs Alloc/Free ............... Ok
Alloc/Insert/Free ............
0001: Adding ["key-0001" / 0x53e18d0] Ok
0002: Adding ["key-0002" / 0x53e1980] Ok
0003: Adding ["key-0003" / 0x53e1a30] Ok
0004: Adding ["key-0004" / 0x53e1ae0] Ok
0005: Adding ["key-0005" / 0x53e1b90] Ok
Ok
==23759==
==23759== HEAP SUMMARY:
==23759== in use at exit: 0 bytes in 0 blocks
==23759== total heap usage: 25 allocs, 25 frees, 1,702 bytes allocated
==23759==
==23759== All heap blocks were freed -- no leaks are possible
==23759==
==23759== For counts of detected and suppressed errors, rerun with: -v
==23759== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

but if I ask 6 items I got SIGSEGV as in:
[email protected] 10907> ./issue 6
libdynamic map testsuite for checking memory with valgrind
Objs Alloc/Free ............... Ok
Alloc/Insert/Free ............
0001: Adding ["key-0001" / 0x7fb3c59255f0] Ok
0002: Adding ["key-0002" / 0x7fb3c59255a0] Ok
0003: Adding ["key-0003" / 0x7fb3c5925550] Ok
0004: Adding ["key-0004" / 0x7fb3c5925500] Ok
0005: Adding ["key-0005" / 0x7fb3c59254b0] Ok
Segmentation fault

and valgrind reports:
==23765== Memcheck, a memory error detector
==23765== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==23765== Using Valgrind-3.12.0 and LibVEX; rerun with -h for copyright info
==23765== Command: ./issue 6
==23765==
libdynamic map testsuite for checking memory with valgrind
Objs Alloc/Free ............... Ok
Alloc/Insert/Free ............
0001: Adding ["key-0001" / 0x53e1980] Ok
0002: Adding ["key-0002" / 0x53e1a30] Ok
0003: Adding ["key-0003" / 0x53e1ae0] Ok
0004: Adding ["key-0004" / 0x53e1b90] Ok
0005: Adding ["key-0005" / 0x53e1c40] Ok
==23765== Invalid read of size 8
==23765== at 0x108CDF: my_equal (issue.c:120)
==23765== by 0x1093B5: map_at (map.c:120)
==23765== by 0x109444: map_insert (map.c:133)
==23765== by 0x108E0C: alloc_insert_free (issue.c:140)
==23765== by 0x108F25: main (issue.c:171)
==23765== Address 0xe is not stack'd, malloc'd or (recently) free'd
==23765==
==23765==
==23765== Process terminating with default action of signal 11 (SIGSEGV)
==23765== Access not within mapped region at address 0xE
==23765== at 0x108CDF: my_equal (issue.c:120)
==23765== by 0x1093B5: map_at (map.c:120)
==23765== by 0x109444: map_insert (map.c:133)
==23765== by 0x108E0C: alloc_insert_free (issue.c:140)
==23765== by 0x108F25: main (issue.c:171)
==23765== If you believe this happened as a result of a stack
==23765== overflow in your program's main thread (unlikely but
==23765== possible), you can try to increase the size of the
==23765== main thread stack using the --main-stacksize= flag.
==23765== The main thread stack size used in this run was 8388608.
0006: Adding ["key-0006" / 0x53e1cf0] ==23765==
==23765== HEAP SUMMARY:
==23765== in use at exit: 474 bytes in 15 blocks
==23765== total heap usage: 29 allocs, 14 frees, 1,804 bytes allocated
==23765==
==23765== LEAK SUMMARY:
==23765== definitely lost: 0 bytes in 0 blocks
==23765== indirectly lost: 0 bytes in 0 blocks
==23765== possibly lost: 0 bytes in 0 blocks
==23765== still reachable: 474 bytes in 15 blocks
==23765== suppressed: 0 bytes in 0 blocks
==23765== Reachable blocks (those to which a pointer was found) are not shown.
==23765== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==23765==
==23765== For counts of detected and suppressed errors, rerun with: -v
==23765== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
Segmentation fault

I suspect the map implementation invoke the callback with a value rather than a pointer.

I would like to have a map of string keys and val pointers to robj_t but I am not sure what is the best way to set the map_construct() parameters and the callbacks.

Also I have experimented unterminated loops in map_at() in the event of a bad return code from my_equal().

Please find attacched my simple test issue example.

/rocco

issue.c.gz