Mawk and mysql conflict

[uzsolt]

[root@pld-laptop ~]# service mysql restart
Error: datadir specified in /var/lib/mysql/mysqld.conf should be
/var/lib/mysql/mysqldb/db MySQL can't be run.

[root@pld-laptop ~]# grep -v "^#" /etc/mysql/clusters.conf

mysqld.conf=/var/lib/mysql

[root@pld-laptop ~]# rpm -qv mysql
mysql-5.1.52-1.i686

[root@pld-laptop ~]# rpm -qv mawk
mawk-1.3.4-0.20100625.1.i686

mawk build on carme, but some time ago there were in th-foo (I don't
know, in which th-*), and --upgrade-dist didn't "downgrade" mawk.

[Caleb Maclennan (alerque)]

What do you get for `grep datadir /var/lib/mysql/mysqld.conf`?

[Artur Frysiak (wiget)]

[root@tower ~]# grep datadir /var/lib/mysql/mysqld.conf
datadir = /var/lib/mysql/mysqldb/db
[root@tower ~]# /etc/init.d/mysql restart
Error: datadir specified in /var/lib/mysql/mysqld.conf should be /var/lib/mysql/mysqldb/db
 MySQL can't be run.

[root@tower ~]# rpm -q mysql mawk gawk
mysql-5.1.52-1.x86_64
mawk-1.3.4-0.20100625.1.x86_64
gawk-3.1.8-1.x86_64

[Caleb Maclennan (alerque)]

Artur: First of all the 1.3.4 package version you have installed is NOT in the th tree, you must have compiled that yourself. It is still marked with a 0.x release number. If you want a working system you should do something like `poldek -n th --upgrade mawk --downgrade` to get the latest stable version in the th tree even though that will be a step down version from your currently installed one.

I have compiled and confirmed the same problem. For others reference here is a relevant line from an strace of the service restart command:

    [pid 22234] execve("/bin/awk", ["awk", "-F=", "\n\t{\n\t\t# undos\n\t\tgsub(/\\r$/, \"\");"..., "/var/lib/mysql/mysqld.conf"], [/* 7 vars */]) =0

[Caleb Maclennan (alerque)]

Confirmed existence of conflict but mawk version in question is an unstable snapshot so fix is low priority.

[Artur Frysiak (wiget)]

Fixed in 5.1.52-3

[Elan Ruusamäe (glen666)]

  On 13.12.2010 17:09, wiget wrote:
> Author: wiget Date: Mon Dec 13 15:09:16 2010 GMT
> Module: packages Tag: HEAD
> ---- Log message:
> - allow spaces and tabs after variable name
...
> section == "mysqld" {
> - if ($1 == "datadir") {
> + if ($1 ~ /^datadir[ \t]*$/) {
> printf("MYSQL_DATA_DIR=%s;", $2);
> - } else if ($1 == "user") {
> + } else if ($1 ~ /^user[ \t]*$/) {
> printf("MYSQL_USER=%s;", $2);
> - } else if ($1 == "pid-file") {
> + } else if ($1 ~ /^pid-file[ \t]*$/) {
> printf("MYSQL_PIDFILE=%s;", $2);
> - } else if ($1 == "socket") {
> + } else if ($1 ~ /^socket[ \t]*$/) {
> printf("MYSQL_SOCKET=%s;", $2);
> - } else if ($1 == "port") {
> + } else if ($1 ~ /^port[ \t]*$/) {
> printf("MYSQL_PORT=%s;", $2);
> - } else if ($1 == "bind-address") {
> + } else if ($1 ~ /^bind-address[ \t]*$/) {
> printf("MYSQL_BIND_ADDRESS=%s;", $2);
> - } else if ($1 == "skip-networking") {
> + } else if ($1 ~ /^skip-networking[ \t]*$/) {
> printf("MYSQL_SKIP_NETWORKING=1;");
> - } else if ($1 == "log-error") {
> + } else if ($1 ~ /^log-error[ \t]*$/) {
> printf("MYSQL_LOG_ERROR=%s;", $2);
> }
>
...pff, make temp var "key" there, which has the spaces/tabs stripped,
instead!

--
glen

[Artur Frysiak (wiget)]

2010/12/13 Elan Ruusamäe <email address hidden>:
>  On 13.12.2010 17:09, wiget wrote:
>>
>> Author: wiget                        Date: Mon Dec 13 15:09:16 2010 GMT
>> Module: packages                      Tag: HEAD
>> ---- Log message:
>> - allow spaces and tabs after variable name
>
> ...
>>
>>        section == "mysqld" {
>> -               if ($1 == "datadir") {
>> +               if ($1 ~ /^datadir[ \t]*$/) {
>>                        printf("MYSQL_DATA_DIR=%s;", $2);
>> -               } else if ($1 == "user") {
>> +               } else if ($1 ~ /^user[ \t]*$/) {
>>                        printf("MYSQL_USER=%s;", $2);
>> -               } else if ($1 == "pid-file") {
>> +               } else if ($1 ~ /^pid-file[ \t]*$/) {
>>                        printf("MYSQL_PIDFILE=%s;", $2);
>> -               } else if ($1 == "socket") {
>> +               } else if ($1 ~ /^socket[ \t]*$/) {
>>                        printf("MYSQL_SOCKET=%s;", $2);
>> -               } else if ($1 == "port") {
>> +               } else if ($1 ~ /^port[ \t]*$/) {
>>                        printf("MYSQL_PORT=%s;", $2);
>> -               } else if ($1 == "bind-address") {
>> +               } else if ($1 ~ /^bind-address[ \t]*$/) {
>>                        printf("MYSQL_BIND_ADDRESS=%s;", $2);
>> -               } else if ($1 == "skip-networking") {
>> +               } else if ($1 ~ /^skip-networking[ \t]*$/) {
>>                        printf("MYSQL_SKIP_NETWORKING=1;");
>> -               } else if ($1 == "log-error") {
>> +               } else if ($1 ~ /^log-error[ \t]*$/) {
>>                        printf("MYSQL_LOG_ERROR=%s;", $2);
>>                }
>>
>
> ...pff, make temp var "key" there, which has the spaces/tabs stripped,
> instead!

glen, you have write access. So, use your rights and write better
code. This "fix" work for me and other (un)happy mysql/mawk users.

--
Artur Frysiak

[Elan Ruusamäe (glen666)]

 On Tue, 14 Dec 2010 00:01:19 +0100, Artur Frysiak wrote:
> 2010/12/13 Elan Ruusamäe <email address hidden>:
>>  On 13.12.2010 17:09, wiget wrote:
>>>
>>> Author: wiget                        Date: Mon Dec 13 15:09:16 2010
>>> GMT
>>> Module: packages                      Tag: HEAD
>>> ---- Log message:
>>> - allow spaces and tabs after variable name
>>
>> ...pff, make temp var "key" there, which has the spaces/tabs
>> stripped,
>> instead!
>
> glen, you have write access. So, use your rights and write better
> code. This "fix" work for me and other (un)happy mysql/mawk users.

 no, fix yourself, i have no testcase, test environ to test if my
 changes fix or break.

 if you refuse to do it sanely, i rather revert your changes, if you
 want to play.

--
 glen

[Caleb Maclennan (alerque)]

>>> ...pff, make temp var "key" there
>> you have write access.
>  no, fix yourself,

Hey guys don't forget we're on the same team here.

I tried making this change myself, but either my smattering of awk
knowledge is not enough or something else is going on here. I made up
a variable and striped the spaces from it and used that for the tests,
but it gives the same error we started with. Then I realized that the
original variable was already being stripped (see line 199)! The ~
comparison must be doing something other than just stripping spaces as
per the regex.

Anybody know what else is going on there? Here is my attempted patch.

Index: mysql.init
===================================================================
RCS file: /cvsroot/packages/mysql/mysql.init,v
retrieving revision 1.144
diff -u -r1.144 mysql.init
--- mysql.init 13 Dec 2010 15:09:11 -0000 1.144
+++ mysql.init 14 Dec 2010 09:24:43 -0000
@@ -209,22 +209,25 @@
                next;
        }

+ key = $1;
+ gsub(/[ \t]*/, "", $key);
+
        section == "mysqld" {
- if ($1 ~ /^datadir[ \t]*$/) {
+ if ($key == "datadir") {
                        printf("MYSQL_DATA_DIR=%s;", $2);
- } else if ($1 ~ /^user[ \t]*$/) {
+ } else if ($key == "user") {
                        printf("MYSQL_USER=%s;", $2);
- } else if ($1 ~ /^pid-file[ \t]*$/) {
+ } else if ($key == "pid-file") {
                        printf("MYSQL_PIDFILE=%s;", $2);
- } else if ($1 ~ /^socket[ \t]*$/) {
+ } else if ($key == "socket") {
                        printf("MYSQL_SOCKET=%s;", $2);
- } else if ($1 ~ /^port[ \t]*$/) {
+ } else if ($key == "port") {
                        printf("MYSQL_PORT=%s;", $2);
- } else if ($1 ~ /^bind-address[ \t]*$/) {
+ } else if ($key == "bind-address") {
                        printf("MYSQL_BIND_ADDRESS=%s;", $2);
- } else if ($1 ~ /^skip-networking[ \t]*$/) {
+ } else if ($key == "skip-networking") {
                        printf("MYSQL_SKIP_NETWORKING=1;");
- } else if ($1 ~ /^log-error[ \t]*$/) {
+ } else if ($key == "log-error") {
                        printf("MYSQL_LOG_ERROR=%s;", $2);
                }
        }

[Elan Ruusamäe (glen666)]

  On 14.12.2010 12:02, Caleb Maclennan wrote:
> + key = $1;
> + gsub(/[ \t]*/, "", $key);
> +
try:

gsub(/[ \t]*/, "", key);

or rather to use the original code:

gsub(/^[\t ]*|[\t ]*$/, "", key);

my suspection is that the first { } block is not invoked in that buggy mawk

--
glen

[Elan Ruusamäe (glen666)]

  On 14.12.2010 12:02, Caleb Maclennan wrote:
 > + if ($key == "datadir") {
 > printf("MYSQL_DATA_DIR=%s;", $2);
and here too:

if (key == "datadir") {

(as the original code was)

--
glen

[Caleb Maclennan (alerque)]

Thanks for your suggestions Elan. I tried them over here but unfortunately I still get the same original error. I've played with it a dozen ways now and what awk is doing with this code defies my imagination. Going back to Artur's code works, but any change to that and it throws up. Simply doing something as simple as substituting $1 for a copy of the variable make it die. Not explicitly matching for the end of string makes it die. Taking too big a bite of chocolate while debugging makes it die.

Frankly after failing to make awk behave the whole stupid eval/awk gig inside this init script is starting to look like a crazy hack. Thoughts?

[Artur Frysiak (wiget)]

glen, please take a look at http://cvs.pld-linux.org/cgi-bin/cvsweb.cgi/packages/mysql/mysql.init.diff?r1=1.144;r2=1.144.2.1

Are you happy with this change?

[Elan Ruusamäe (glen666)]

yes absolutely, i assume it works and sent new rel to th

[Artur Frysiak (wiget)]

Caleb and uzsolt , please test 5.1.52-4 and let me know if it works for you.

[uzsolt (udvzsolt)]

Works for me with mawk-1.3.4-0.20100625.1.i686.

[Elan Ruusamäe (glen666)]

adding test script to dest mawk

$ q mawk gawk
mawk-1.3.3-33.x86_64
gawk-3.1.8-1.x86_64

$ gawk -f 689563.awk
ORIGINAL: [pff]
TRIM1: [pff]
TRIM2: [pff]
ORIGINAL: [ leading-space]
TRIM1: [leading-space]
TRIM2: [leading-space]
ORIGINAL: [trailing-space ]
TRIM1: [trailing-space]
TRIM2: [trailing-space]
ORIGINAL: [ leading-tab]
TRIM1: [leading-tab]
TRIM2: [leading-tab]
ORIGINAL: [traling-tab ]
TRIM1: [traling-tab]
TRIM2: [traling-tab]

$ mawk -f 689563.awk
ORIGINAL: [pff]
TRIM1: [pff]
TRIM2: [pff]
ORIGINAL: [ leading-space]
TRIM1: [leading-space]
TRIM2: [leading-space]
ORIGINAL: [trailing-space ]
TRIM1: [trailing-space]
TRIM2: [trailing-space]
ORIGINAL: [ leading-tab]
TRIM1: [leading-tab]
TRIM2: [leading-tab]
ORIGINAL: [traling-tab ]
TRIM1: [traling-tab]
TRIM2: [traling-tab]
$

[Caleb Maclennan (alerque)]

Good to go here Artur, thanks to all and grr to awk!