generate UTF locales selectable in icingaweb2

[skm42]

#101

[jjethwa]

Hi @skm42

Thanks so much for the PR! I think it might be better to move this step to the container run script to reduce layer size? I guess we need to see how it interacts with: #141 as well

[skm42]

trade-off between image size and container initialization time, I don't mind. like this? skm42@5700e23

Concerning #141 I took a look into icinga2 standalone vagrant box https://github.com/Icinga/icinga-vagrant. It has all locales generated, utf and iso ones. Should be save to follow like this, what's your view? skm42@aaeeebb

[bebehei]

Thanks so much for the PR! I think it might be better to move this step to the container run script to reduce layer size?

Sorry, but this is pointless. It's absolutely static and not even parametrized. You buy this with container boottime, which is quite valuable on container reboot.

The locales package is ~15 MBs big and all generated locales are about 6MBs in size. Compared to an image, which is about 600MBs big ....

---

@skm42 Please put the stuff back again into the Dockerfile. But do you see the ARGs in the file?

The stuff you want to add here is currently completely static and has no parameters. Please pack it above the ARGs, as changing ARGs are invalidating the cache and require this layer to get rebuilt, albeit nothing will be different in this layer.

Please squash the commits, too and have a look at the en_US locale it looks like a typo as it doesn't include the actual the charset.

---

Concerning #141 I took a look into icinga2 standalone vagrant box https://github.com/Icinga/icinga-vagrant. It has all locales generated, utf and iso ones.

Could you please provide a direct link for this? I'm not able to find the part you may refer to.

[skm42]

Please see latest commit.
Keep locales to build configurable by ENV: BUILD_LOCALE="locale 1,locale 2"
Locale has to be defined following /etc/locale.gen
What needs to be improved is how to identify whether a locale is already build.

root@icinga2:/# locale -a |grep -i en_us.utf
en_US.utf8
root@icinga2:/# cat /etc/locale.gen |grep -i en_us.utf
en_US.UTF-8 UTF-8


Tested for utf locales ar_SA.UTF-8 UTF-8,de_DE.UTF-8 UTF-8,en_US.UTF-8,fi_FI.UTF-8 UTF-8,it_IT.UTF-8 UTF-8,pt_BR.UTF-8 UTF-8,ru_RU.UTF-8 UTF-8, others not yet.

[bebehei]

Sorry, I'm laughing about your code. It looks like a you're a proper C programmer, miss all C idioms and want to implement them in bash 😂

Your code looks like a poem written in swollen language, which tries to express the meaning of a simple sentence.

But in contrast to a poem, you have to write down the simple sentence here. People need to understand you and not have to interpret your peom, what it could mean.

It seems that you had a pretty hard job in programming bash/shell. I guess, I can help you, by commenting your code and showing you, what could get improved.

I'm going to reject this PR, as I've found a welldoing solution.

[bebehei]

There is no guarantee, that your process will spit out a real number. So typesetting it to an integer doesn't give you any guarantee for this. Also quote the commands output, as it could contain special shell chars or just spaces, which get interpreted.

[bebehei]

grep -q "^${b_locale}" /etc/locale.gen

[bebehei]

You're iterating here over an index. Well, that's possible, but in bash this is discouraged. The language is not made for this. Use things like an enhanced for loop.

for elem in $(command_which spits out elements); do
   # do some stuff with "${elem}"
done

Rule of thumb: At the point where you have to use real integers either

• bash isn't the way to go and another language fits much better
• you misuse bash
• you know that you misuse bash, but it's the only language available

[bebehei]

The if clause actually executes a command and then tests for its exit code. If the exit code is 0 (==true), it'll execute the codeblock the result else not.

So the [[ (or [ depending on your shell) is called the test function, which is actually a just regular command, getting fed with the logical operators and evaluating them. The "matching" ]] isn't actually a bash syntactic token, it's just an argument to the command and this command can verify that it received all arguments.

So, extratesting for $? -ne 0 is superfluous, as you just could use if grep -q "^$b_locale" /etc/locale.gen; then and three lines are shrunken to a single one.

[bebehei]

If the locale, I specified is not there in the file, it won't get enabled. On the one hand side, this is great. But KIM, that you could have a typo in the environment variable (e.g.: using en_us.UTF-8 UTF-8) and then the actual locale gets silently dropped.

---

sed has got a -i switch. And also like grep it takes a filename as the last parameters.

sed -i "s/# $b_locale/$b_locale/" /etc/locale.gen

[bebehei]

Not further tested, but this would basically do the same:

echo "${BUILD_LOCALES}" | while IFS=, read locale charset; do
    sed -i "s/#\\s*${locale}/${locale}/" /etc/locale.gen
done

[bebehei]

echo "${BUILD_LOCALES}" | while IFS=, read locale charset; do
    if locale -a | grep -qi "^${locale}$"; do
        locale-gen
        break
    fi
done

[skm42]

fine for me. Going to review your comments for learning better bash and maybe come up with a new PR at a later time.

[bebehei]

@skm42 Thanks for firing up this PR. You forced me to finally fix it. #142

[jjethwa]

Thanks @bebehei and @skm42 Looks like it was a good brainstorming session while I was away 😛