# Openresty downstream bundle maintenance requests

**URL:** <https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927>\
**Category:** NGINX\
**Created:** [September 17, 2011, 10:53am UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927 "2011-09-17T10:53:59Z")\
**Posts on this page:** 9\
**Page:** 1

<div class="post-metadata">

**Author:** ![dubstep](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/d/cdc98d/32.png) [@dubstep](https://www.ruby-forum.com/u/dubstep)\
**Post date:** [September 17, 2011, 10:53am UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/1 "2011-09-17T10:53:59Z")

</div>

Hi.

A few things to consider with respect to downstream maintenance support.

The configure script under the “for my $opt (@ARGV) {” loop covers a few  
typical items and any others cause the “die “Invalid option $opt\n”;”  
line  
to be triggered. However, when in a Fedora/Enterprise Linux build  
environment, the “%configure” macro in the spec file adds several other  
switches (which work fine when building nginx directly).

Can the upstream please consider modifying the configure script to  
allow  
for these?

A possiblity is:

for my $opt (@ARGV) {  
…

```
} elsif ($opt =~ /^--build=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--host=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--target=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--program-prefix=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--exec-prefix=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--bindir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--sbindir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--sysconfdir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--datadir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--includedir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--libexecdir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--localstatedir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--sharedstatedir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--mandir=(.*)/) {
    push @ngx_opts, $opt;

} elsif ($opt =~ /^--infodir=(.*)/) {
    push @ngx_opts, $opt;

} else {
    die "Invalid option $opt\n";
}

```

}

May be even merge all the “dir” ones into “} elsif ($opt =~  
/^–(._)dir=(._)/) {”

Also, can the upstream consider dropping version numbers from the  
folders in  
the “bundle” directory and record the version numbers in a README? Would  
save having to edit the spec file for every release.

Finally, can the upstream consider allowing the with luajit option to be  
selected but not to try to build luajit? I.E. set openresty up to use  
luajit  
but have the luajit build done separately and not try to replace this by  
the  
standard lua. I suppose the same applies to the standard lua. Basically,  
third party applications should ideally be built separately and we would  
look into creating seperate rpms for those. Can foresee some potential  
knotty issues to overcome with this but if distributing an rpm, then  
there  
are issues anyway.

Thanks

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 17, 2011, 3:58pm UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/2 "2011-09-17T15:58:53Z")

</div>

On Sat, Sep 17, 2011 at 4:53 PM, Nginx U. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> The configure script under the “for my $opt (@ARGV) {” loop covers a few  
> typical items and any others cause the “die “Invalid option $opt\n”;” line  
> to be triggered. However, when in a Fedora/Enterprise Linux build  
> environment, the “%configure” macro in the spec file adds several other  
> switches (which work fine when building nginx directly).

Which options do not work for you? The last part in that loop is  
actually:

```
} elsif ($opt =~ /^--\w.*/) {
    push @ngx_opts, $opt;

} else {
    die "Invalid option $opt\n";
}

```

That is, any options starting with a /\w/ will be matched in the last  
2nd condition and passed to nginx’s configure script.

> ```
> push @ngx_opts, $opt;
> 
> ```
> 
> } elsif ($opt =~ /^–datadir=(._)/) {  
> push @ngx\_opts, $opt;  
> } elsif ($opt =~ /^–infodir=(._)/) {  
> push @ngx\_opts, $opt;  
> } else {  
> die “Invalid option $opt\n”;  
> }  
> }

These options should be happily accepted by the last 2nd option. If  
not, please let me know 🙂

> Also, can the upstream consider dropping version numbers from the folders in  
> the “bundle” directory and record the version numbers in a README? Would  
> save having to edit the spec file for every release.

Okay, I’ll remove version numbers in my build script in the next  
release of ngx\_openresty 🙂

> Finally, can the upstream consider allowing the with luajit option to be  
> selected but not to try to build luajit? I.E. set openresty up to use luajit  
> but have the luajit build done separately and not try to replace this by the  
> standard lua. I suppose the same applies to the standard lua. Basically,  
> third party applications should ideally be built separately and we would  
> look into creating seperate rpms for those.

Will adding new options like --with-luajit=PATH and --with-lua51=PATH  
work for you? That is, telling ngx\_openresty to use an external luajit  
or lua 5.1 build located by PATH?

By default, ngx\_openresty will continue try to build its bundled  
version of LuaJIT or Lua 5.1 and use it whenever possible and that’s  
the whole point of a bundle 😉 But I’m still willing to add special  
options to help RPM/Debian/etc packaging for sure 🙂

> Can foresee some potential  
> knotty issues to overcome with this but ifdistributingan rpm, then there  
> are issues anyway.

I’m very willing to help and solve any issues along the way 😉

Thanks!  
-agentzh

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 17, 2011, 4:08pm UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/3 "2011-09-17T16:08:45Z")

</div>

On Sat, Sep 17, 2011 at 4:53 PM, Nginx U. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> Also, can the upstream consider dropping version numbers from the folders in  
> the “bundle” directory and record the version numbers in a README? Would  
> save having to edit the spec file for every release.

Just out of curiosity, why will you hard-code the directory names  
under bundle/ in your rpm spec file?

Regards,  
-agentzh

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 17, 2011, 4:43pm UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/4 "2011-09-17T16:43:56Z")

</div>

On Sat, Sep 17, 2011 at 4:58 PM, agentzh [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> } elsif ($opt =~ /^–\w.\*/) {  
> push @ngx\_opts, $opt;
> 
> } else {  
> die “Invalid option $opt\n”;  
> }
> 
> That is, any options starting with a /\w/ will be matched in the last  
> 2nd condition and passed to nginx’s configure script.

Correct and it matches all the “–with XYZ” and “–without XYZ” flags  
for  
nginx.

> > ```
> > push @ngx_opts, $opt;
> > } elsif ($opt =~ /^--datadir=(.*)/) {
> > push @ngx_opts, $opt;
> > } elsif ($opt =~ /^--infodir=(.*)/) {
> > push @ngx_opts, $opt;
> > } else {
> > die "Invalid option $opt\n";
> > }
> > 
> > ```
> > 
> > }
> 
> These options should be happily accepted by the last 2nd option. If  
> not, please let me know 🙂

They are matched by the last option of ‘die “Invalid option $opt\n”;’ .

> > Also, can the upstream consider dropping version numbers from the folders  
> > in  
> > the “bundle” directory and record the version numbers in a README? Would  
> > save having to edit the spec file for every release.
> 
> Okay, I’ll remove version numbers in my build script in the next  
> release of ngx\_openresty 🙂

No longer required thanks.

> Will adding new options like --with-luajit=PATH and --with-lua51=PATH  
> work for you? That is, telling ngx\_openresty to use an external luajit  
> or lua 5.1 build located by PATH?

This should work. We can then switch piar this with --without luajit. We  
obviously want to keep the lua-rds-parser and lua-redis options.

> I’m very willing to help and solve any issues along the way 😉
> 
> Thanks!  
> -agentzh

Thanks!

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 17, 2011, 5:22pm UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/5 "2011-09-17T17:22:13Z")

</div>

On Sat, Sep 17, 2011 at 10:43 PM, Nginx U. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid)  
wrote:

> > These options should be happily accepted by the last 2nd option. If  
> > not, please let me know 🙂
> 
> They are matched by the last option of ‘die “Invalid option $opt\n”;’ .

I’ve just confirmed that at least “–includedir=PATH” and  
“–datadir=PATH” are accepted by ngx\_openresty’s configure script but  
rejected by nginx’s configure script:

```
~/work/nginx-1.0.6 $ ./configure --includedir=/tmp/blah
./configure: error: invalid option "--includedir=/tmp/blah"

~/work/nginx-1.0.6 $ ./configure --datadir=/tmp/blah
./configure: error: invalid option "--datadir=/tmp/blah"

```

Please note that the ‘error: invalid option “.\*”’ messages  
_different_ from the one output by ‘die “Invalid option $opt\n”;’.

Regards,  
-agentzh

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 18, 2011, 7:35am UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/6 "2011-09-18T07:35:10Z")

</div>

On Sat, Sep 17, 2011 at 10:43 PM, Nginx U. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid)  
wrote:

> On Sat, Sep 17, 2011 at 4:58 PM, agentzh [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:
> 
> > Will adding new options like --with-luajit=PATH and --with-lua51=PATH  
> > work for you? That is, telling ngx\_openresty to use an external luajit  
> > or lua 5.1 build located by PATH?
> 
> This should work. We can then switch piar this with --without luajit. We  
> obviously want to keep the lua-rds-parser and lua-redis options.

I’ve just added the --with-luajit=PATH and --with-lua51=PATH options  
to ngx\_openresty’s configure script and this update has already been  
included in the latest devel release, 1.0.6.7:

```
http://openresty.org/#Download

```

Please test it out to see if it fits your requirements 🙂 All of those  
Lua modules bundled by ngx\_openresty will still be built by default  
when either --with-luajit=PATH or --with-lua51=PATH is specified  
unless being explicitly disabled.

Enjoy!  
-agentzh

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 18, 2011, 6:19pm UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/7 "2011-09-18T18:19:08Z")

</div>

On Sun, Sep 18, 2011 at 8:34 AM, agentzh [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> Enjoy!  
> -agentzh

Hi.

I’ll have a go later.

Just as an update, please ignore the other requests. The only one that  
needed attention was this lua related one.

Thanks

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 18, 2011, 11:08am UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/8 "2011-09-18T11:08:07Z")

</div>

On Sat, Sep 17, 2011 at 6:42 PM, Nginx U. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> > > They are matched by the last option of ‘die “Invalid option $opt\n”;’ .  
> > > ./configure: error: invalid option “–datadir=/tmp/blah”
> 
> However things crash with the openresty configure as said. Specifically,  
> the message returned is for on “–build” which is the first one encountered.  
> Recreating the build machine at the moment so can’t give specifics right  
> now. I’ll try changing the text in ‘die “Invalid option $opt\n”;’ when  
> having another go to verify whether it is actually this or not. Looking at  
> the code though, it seems clear that these extra options will fail at the  
> specified point.
> 
> Thanks.

Hi.

Just looking at the spec file example I linked to, I can see it is not  
using  
the rpm %configure macro. So you were on to something there.

I suppose the only remaining valid request from my original post is the  
lua  
one.

Thanks

---

<div class="post-metadata">

**Author:** ![Nginx\_U](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/n/dec6dc/32.png) [@Nginx\_U](https://www.ruby-forum.com/u/Nginx_U)\
**Post date:** [September 18, 2011, 11:31pm UTC](https://www.ruby-forum.com/t/openresty-downstream-bundle-maintenance-requests/211927/9 "2011-09-18T23:31:58Z")

</div>

On Sat, Sep 17, 2011 at 11:42 PM, Nginx U. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid)  
wrote:

> +# This combined with Fedora’s FORTIFY\_SOURCE=2 option causes it nginx  
> +# to not compile.  
> +#CFLAGS=“$CFLAGS -Werror”
> 
> # debug
> 
> CFLAGS=“$CFLAGS -g”

The nginx core bundled by ngx\_openresty has this patch as well.

> The standard outputs from the rpm %configure macro including all those  
> elements build fine after this when building nginx by itself. An example of  
> the typical spec file with this patch  
> ishttp://www4.atomicorp.com/channels/source/nginx/nginx.spec (they are not  
> up to date but it does work as is for 1.0.6 and should for the 1.1.x branch  
> as well.  
> However things crash with the openresty configure as said. Specifically, the  
> message returned is for on “–build” which is the first one encountered.

The configure command in the RPM spec file you link to does not crash  
on my slackware linux x86\_64 for ngx\_openresty 1.0.6.5. I wonder what  
error messages are you getting?

> Recreating the build machine at the moment so can’t give specifics right  
> now. I’ll try changing the text in’die “Invalid option $opt\n”;’ when  
> having another go to verify whether it is actually this or not. Looking at  
> the code though, it seems clear that these extra options will fail at the  
> specified point.

Why won’t “–build” be matched by /^–\w.\*/ in ngx\_openresty’s  
configure script? I’ve said extra options would be captured by the  
last 2nd condition and be passed verbatim to nginx’s configure script  
and would not reach the last “die” statement there.

Regards,  
-agentzh