To compile OQGraph v2 you need to have the boost library with the versions not earlier than 1.40 and not later than 1.55
but lets ask andymc73 if it's still supposed to be the case.
Elena Stepanova
added a comment - It is actually documented here: https://mariadb.com/kb/en/mariadb/compiling-oqgraph/
To compile OQGraph v2 you need to have the boost library with the versions not earlier than 1.40 and not later than 1.55
but lets ask andymc73 if it's still supposed to be the case.
To compile OQGraph v2 you need to have the boost library with the versions not earlier than 1.40 and not later than 1.55
but lets ask Andrew McDonnell if it's still supposed to be the case.
I believe MariaDB 10 and above use OQGraph v3. So it is a valid bug
Brian Evans
added a comment - @elenst
It is actually documented here: https://mariadb.com/kb/en/mariadb/compiling-oqgraph/
To compile OQGraph v2 you need to have the boost library with the versions not earlier than 1.40 and not later than 1.55
but lets ask Andrew McDonnell if it's still supposed to be the case.
I believe MariaDB 10 and above use OQGraph v3. So it is a valid bug
There is no "v3" - ha_ograph.cc reports "3.0" but this would seem to be a very long standing typo. As things stand, the wiki is correct, until someone submits a patch to make it work with newer systems
v1 was the original prototype by Antony which TTBOMK used a heavy reliance on the internals of the then MySQL 5.1
v2 was a rewrite using boost graph and a more correct storage engine implementation and virtual, which is what currently exists. Again TTBOMK, this was all first built before I got everything working for MariaDB 10.0
Andrew McDonnell
added a comment - - edited There is no "v3" - ha_ograph.cc reports "3.0" but this would seem to be a very long standing typo. As things stand, the wiki is correct, until someone submits a patch to make it work with newer systems
v1 was the original prototype by Antony which TTBOMK used a heavy reliance on the internals of the then MySQL 5.1
v2 was a rewrite using boost graph and a more correct storage engine implementation and virtual, which is what currently exists. Again TTBOMK, this was all first built before I got everything working for MariaDB 10.0
> MariaDB starting with 10.0.7
> OQGraph v3 compiles fine with the newer boost libraries, but it additionally needs Judy library installed.
And, for example, I do have judy installed, and had mariadb-10.1.12 built fine against boost-1.58. But now mariadb-10.1.13 refuses to build against boost-1.60 with the error, mentioned by bug reporter. So, I'd say it is some specific boost-1.60 incompatibility, or bulding against 1.58 with judy installed was a coincedence...
Vadim A. Misbakh-Soloviov
added a comment - But wiki also states:
> MariaDB starting with 10.0.7
> OQGraph v3 compiles fine with the newer boost libraries, but it additionally needs Judy library installed.
And, for example, I do have judy installed, and had mariadb-10.1.12 built fine against boost-1.58. But now mariadb-10.1.13 refuses to build against boost-1.60 with the error, mentioned by bug reporter. So, I'd say it is some specific boost-1.60 incompatibility, or bulding against 1.58 with judy installed was a coincedence...
Dimitri John Ledkov
added a comment - Is https://github.com/boostorg/graph/commit/e77cd87de11263b60cebf6996efc7d1b3cb4d0b5#diff-3d8548656a5db728e028023e3f803acdL385 of any help?
For reference, this comment and the follow-up comment on an older bug are probably relevant. In summary, yes there is a v3, and it has different boost requirements than v2.
FWIW, I have mariadb 10.0.24 here compiled against boost 1.59.0 and all is good. So it's most likely something introduced in 1.60.0.
Heinz Wiesinger
added a comment - For reference, this comment and the follow-up comment on an older bug are probably relevant. In summary, yes there is a v3, and it has different boost requirements than v2.
FWIW, I have mariadb 10.0.24 here compiled against boost 1.59.0 and all is good. So it's most likely something introduced in 1.60.0.
Looked into this and it appears the typedef's are gone in Boost 1.6. There might be an easy fix for this, which I'm attempting now.
Vicențiu Ciorbaru
added a comment - Looked into this and it appears the typedef's are gone in Boost 1.6. There might be an easy fix for this, which I'm attempting now.
Vicențiu Ciorbaru
added a comment - Hi Sergei,
Can you please review the patch for this bug?
https://lists.launchpad.net/maria-developers/msg09673.html
cvicentiu Please review that patch and apply something from there is its good.
Otto Kekäläinen
added a comment - I just noticed that an Ubuntu dev patched the 10.0.24-7 package in Yaketty (without consulting upstream) to support Boost 1.6. His patch is visible here: https://patches.ubuntu.com/m/mariadb-10.0/mariadb-10.0_10.0.24-7ubuntu1.patch
cvicentiu Please review that patch and apply something from there is its good.
It is functionally equivalent to what was posted on the mailing list, and it's irrelevant which patch is taken =) or something else is done.
Regards,
Dimitri.
Dimitri John Ledkov
added a comment - @Otto
An Ubuntu dev is I, who commented on here https://jira.mariadb.org/browse/MDEV-9479?focusedCommentId=83042&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-83042 and also attached the patch to the same effect to this ticket on 10th of May. I'm not sure if that got missed, or I did something wrong in jira.
It is functionally equivalent to what was posted on the mailing list, and it's irrelevant which patch is taken =) or something else is done.
Regards,
Dimitri.
This also affects MariaDB v10.1.14, when compiled against Boost v1.6.1.0 (On Void Linux, with both MariaDB and Boost compiled from source).
Can confirm Dimitri's patch works (thanks).
Antonio Malcolm
added a comment - This also affects MariaDB v10.1.14, when compiled against Boost v1.6.1.0 (On Void Linux, with both MariaDB and Boost compiled from source).
Can confirm Dimitri's patch works (thanks).
It is actually documented here: https://mariadb.com/kb/en/mariadb/compiling-oqgraph/
but lets ask andymc73 if it's still supposed to be the case.