Package marked as broken, install from Github source?

freejetz888

New Member

Reaction score: 5
Messages: 18

I have a package that fails to build in my Poudriere repository, databases/postgresql-plv8js, due to it being marked as broken. Unfortunately this is a required package for the piece of ERP software the PostgreSQL database is built for. I have found the Github source for plv8 and I'm trying to figure out if it is possible to build from there into my Poudriere repository to distribute to my servers that need it. If not possible would I be able to install directly onto the servers bypassing Poudriere.

I have done some searching and found a couple sources that makes me think this is possible but wonder if I am wasting my time. If anyone could point me in the right direction or manuals it would be greatly appreciated. Thanks!
 
Last edited by a moderator:

dd_ff_bb

Member

Reaction score: 83
Messages: 97

I guess you have 2 options:

1) try to fix the port version. its marked broken with error on 01 Jan 2021 20:58:57 you can see error details from below link:

2) you can download github version and port it to freebsd.

If you can 1st option seems easier
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 12,755
Messages: 39,353

The port uses the same source as that Github page. So I suggest to try and fix the port.
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 12,755
Messages: 39,353

Looking at the error it may just need a simple patch. I would definitely try to change that std::remove_cv_t to std::remove_cv and see if that works.


You can also try just updating it, the port builds version 1.4.8 but on the github page there are many newer versions available (https://github.com/plv8/plv8/tags). So the issue may already have been resolved in the code.
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

I found somewhere on the internet 😏 (so, well, not sure it can be trusted) that std::remove_cv_t was introduced with C++14. If that's true, a USES= compiler:c++14-lang would probably be the better fix.
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 12,755
Messages: 39,353

I found somewhere on the internet
I think we read the same thing, was it on some page about NodeJS? :D

If that's true, a USES= compiler:c++14-lang would probably be the better fix.
Agreed. But I would also recommend checking out a newer version of the code, this issue may already have been fixed, 1.4.8 is pretty old (released in April 2016). Current version appears to be v3.0.0.

 
OP
F

freejetz888

New Member

Reaction score: 5
Messages: 18

Thank you for the replies. I have done some testing this morning, though I am not exactly sure if I am testing it correctly.

So far I've gone in and modified the Makefile located at /usr/local/poudriere/ports/TEST/databases/postgresql-plv8js to comment out the 'Broken' tag and to modify the compiler to:
Code:
USES=    compiler:c++14-lang
This results in the same error so I'm not sure if this was the correct place to make this change. I have also changed in the make file the version from 1.4.8 to 3.0.0 which fails to fetch.

Code:
=> plv8-plv8-v3.0.0_GH0.tar.gz is not in /usr/ports/databases/postgresql-plv8js/distinfo.
=> Either /usr/ports/databases/postgresql-plv8js/distinfo is out of date, or
=> plv8-plv8-v3.0.0_GH0.tar.gz is spelled incorrectly.
*** Error code 1

I've never really done maintenance of ports like this so I could be off the mark here. I uploaded the plv8-3.0.0 that SirDice linked to my Poudriere server but I'm not sure how to tell Poudriere to use that as the source to build the package.

Below is the log output of the failure using the 1.4.8 version with the compiler:c++14-lang:

Code:
===========================================================================
=======================<phase: configure      >============================
===>  Configuring for postgresql10-plv8js-1.4.8_6
===========================================================================
=======================<phase: build          >============================
===>  Building for postgresql10-plv8js-1.4.8_6
gmake[1]: Entering directory '/wrkdirs/usr/ports/databases/postgresql-plv8js/work/plv8-1.4.8'
sed -e 's/^#undef PLV8_VERSION/#define PLV8_VERSION "1.4.8"/' plv8_config.h.in > plv8_config.h
c++ -Wall -O2 -std=c++11 -fno-rtti  -I. -I./ -I/usr/local/include/postgresql/server -I/usr/local/include/postgresql/internal  -I/usr/local/include -I/usr/local/include -I/usr/local/include -fPIC -c -o plv8.o plv8.cc
In file included from plv8.cc:8:
In file included from ./plv8.h:12:
In file included from /usr/local/include/v8.h:30:
/usr/local/include/v8-internal.h:506:38: error: no template named 'remove_cv_t' in namespace 'std'; did you mean 'remove_cv'?
            !std::is_same<Data, std::remove_cv_t<T>>::value>::Perform(data);
                                ~~~~~^~~~~~~~~~~
                                     remove_cv
/usr/include/c++/v1/type_traits:697:50: note: 'remove_cv' declared here
template <class _Tp> struct _LIBCPP_TEMPLATE_VIS remove_cv
                                                 ^
In file included from plv8.cc:8:
./plv8.h:86:6: error: no type named 'ExternalArrayType' in namespace 'v8'
        v8::ExternalArrayType ext_array;
        ~~~~^
./plv8.h:174:41: error: no member named 'NewSymbol' in 'v8::String'
                                        context->Global()->Get(v8::String::NewSymbol("plv8")));
                                                               ~~~~~~~~~~~~^
./plv8.h:180:30: error: too few arguments to function call, expected 2, have 1
                                        v8::External::New(fcinfo));
                                        ~~~~~~~~~~~~~~~~~       ^
/usr/local/include/v8.h:6062:3: note: 'New' declared here
  static Local<External> New(Isolate* isolate, void* value);
  ^
In file included from plv8.cc:8:
./plv8.h:211:40: error: no member named 'NewSymbol' in 'v8::String'
                                context->Global()->Get(v8::String::NewSymbol("plv8")));
                                                       ~~~~~~~~~~~~^
./plv8.h:217:32: error: too few arguments to function call, expected 2, have 1
                                                                        v8::External::New(conv));
                                                                        ~~~~~~~~~~~~~~~~~     ^
/usr/local/include/v8.h:6062:3: note: 'New' declared here
  static Local<External> New(Isolate* isolate, void* value);
  ^
In file included from plv8.cc:8:
./plv8.h:219:36: error: too few arguments to function call, expected 2, have 1
                                                                        v8::External::New(tupstore));
                                                                        ~~~~~~~~~~~~~~~~~         ^
/usr/local/include/v8.h:6062:3: note: 'New' declared here
  static Local<External> New(Isolate* isolate, void* value);
  ^
In file included from plv8.cc:8:
./plv8.h:247:55: error: no type named 'Arguments' in namespace 'v8'
extern v8::Handle<v8::Value> Subtransaction(const v8::Arguments& args) throw();
                                                  ~~~~^
In file included from plv8.cc:19:
In file included from /usr/local/include/postgresql/server/commands/trigger.h:17:
In file included from /usr/local/include/postgresql/server/nodes/execnodes.h:17:
In file included from /usr/local/include/postgresql/server/access/genam.h:19:
In file included from /usr/local/include/postgresql/server/nodes/tidbitmap.h:26:
In file included from /usr/local/include/postgresql/server/utils/dsa.h:17:
In file included from /usr/local/include/postgresql/server/port/atomics.h:70:
/usr/local/include/postgresql/server/port/atomics/arch-x86.h:143:2: warning: 'register' storage class specifier is deprecated and incompatible with C++17 [-Wdeprecated-register]
        register char _res = 1;
        ^~~~~~~~~
In file included from plv8.cc:19:
In file included from /usr/local/include/postgresql/server/commands/trigger.h:17:
In file included from /usr/local/include/postgresql/server/nodes/execnodes.h:31:
In file included from /usr/local/include/postgresql/server/storage/condition_variable.h:25:
/usr/local/include/postgresql/server/storage/s_lock.h:226:2: warning: 'register' storage class specifier is deprecated and incompatible with C++17 [-Wdeprecated-register]
        register slock_t _res = 1;
        ^~~~~~~~~
plv8.cc:229:14: error: no member named 'Dispose' in 'v8::Persistent<v8::Object, v8::NonCopyablePersistentTraits<v8::Object> >'
                        env->recv.Dispose();
                        ~~~~~~~~~ ^
plv8.cc:230:14: error: no member named 'Clear' in 'v8::Persistent<v8::Object, v8::NonCopyablePersistentTraits<v8::Object> >'
                        env->recv.Clear();
                        ~~~~~~~~~ ^
plv8.cc:270:15: error: calling a protected constructor of class 'v8::HandleScope'
                HandleScope     handle_scope;
                                ^
/usr/local/include/v8.h:1233:13: note: declared protected here
  V8_INLINE HandleScope() = default;
            ^
plv8.cc:276:17: error: no matching function for call to 'CreateExecEnv'
                        proc->xenv = CreateExecEnv(proc->cache->function);
                                     ^~~~~~~~~~~~~
plv8.cc:132:23: note: candidate function not viable: no known conversion from 'Persistent<v8::Function>' to 'Handle<v8::Function>' (aka 'Local<v8::Function>') for 1st argument
static plv8_exec_env *CreateExecEnv(Handle<Function> script);
                      ^
plv8.cc:329:17: error: calling a protected constructor of class 'v8::HandleScope'
                HandleScope                     handle_scope;
                                                ^
/usr/local/include/v8.h:1233:13: note: declared protected here
  V8_INLINE HandleScope() = default;
            ^
plv8.cc:332:19: error: no viable conversion from 'Persistent<v8::Context>' to 'Handle<v8::Context>' (aka 'Local<v8::Context>')
                Handle<Context> global_context = GetGlobalContext();
                                ^                ~~~~~~~~~~~~~~~~~~
/usr/local/include/v8.h:201:7: note: candidate constructor (the implicit copy constructor) not viable: no known conversion from 'Persistent<v8::Context>' to 'const v8::Local<v8::Context> &' for 1st argument
class Local {
      ^
/usr/local/include/v8.h:201:7: note: candidate constructor (the implicit move constructor) not viable: no known conversion from 'Persistent<v8::Context>' to 'v8::Local<v8::Context> &&' for 1st argument
/usr/local/include/v8.h:205:13: note: candidate template ignored: could not match 'Local' against 'Persistent'
  V8_INLINE Local(Local<S> that)
            ^
/usr/local/include/v8.h:349:22: note: explicit constructor is not a candidate
  explicit V8_INLINE Local(T* that) : val_(that) {}
                     ^
plv8.cc:373:12: error: no matching constructor for initialization of 'v8::TryCatch'
        TryCatch                try_catch;
                                ^
/usr/local/include/v8.h:10249:12: note: candidate constructor not viable: requires single argument 'isolate', but no arguments were provided
  explicit TryCatch(Isolate* isolate);
           ^
/usr/local/include/v8.h:10371:3: note: candidate constructor not viable: requires 1 argument, but 0 were provided
  TryCatch(const TryCatch&) = delete;
  ^
plv8.cc:377:58: error: too few arguments to function call, expected 4, have 3
        Local<v8::Value> result = fn->Call(receiver, nargs, args);
                                  ~~~~~~~~                      ^
/usr/local/include/v8.h:4779:3: note: 'Call' declared here
  V8_WARN_UNUSED_RESULT MaybeLocal<Value> Call(Local<Context> context,
  ^
/usr/local/include/v8config.h:502:31: note: expanded from macro 'V8_WARN_UNUSED_RESULT'
#define V8_WARN_UNUSED_RESULT __attribute__((warn_unused_result))
                              ^
plv8.cc:393:19: error: no viable conversion from 'Persistent<v8::Context>' to 'Handle<v8::Context>' (aka 'Local<v8::Context>')
        Handle<Context>         context = xenv->context;
                                ^         ~~~~~~~~~~~~~
/usr/local/include/v8.h:201:7: note: candidate constructor (the implicit copy constructor) not viable: no known conversion from 'Persistent<v8::Context>' to 'const v8::Local<v8::Context> &' for 1st argument
class Local {
      ^
/usr/local/include/v8.h:201:7: note: candidate constructor (the implicit move constructor) not viable: no known conversion from 'Persistent<v8::Context>' to 'v8::Local<v8::Context> &&' for 1st argument
/usr/local/include/v8.h:205:13: note: candidate template ignored: could not match 'Local' against 'Persistent'
  V8_INLINE Local(Local<S> that)
            ^
/usr/local/include/v8.h:349:22: note: explicit constructor is not a candidate
  explicit V8_INLINE Local(T* that) : val_(that) {}
                     ^
plv8.cc:421:35: error: member reference type 'Persistent<v8::Object>' is not a pointer; did you mean to use '.'?
                Local<Function>::Cast(xenv->recv->GetInternalField(0));
                                      ~~~~~~~~~~^~
                                                .
plv8.cc:421:37: error: no member named 'GetInternalField' in 'v8::Persistent<v8::Object, v8::NonCopyablePersistentTraits<v8::Object> >'
                Local<Function>::Cast(xenv->recv->GetInternalField(0));
                                      ~~~~~~~~~~  ^
fatal error: too many errors emitted, stopping now [-ferror-limit=]
2 warnings and 20 errors generated.
gmake[1]: *** [Makefile:68: plv8.o] Error 1
gmake[1]: Leaving directory '/wrkdirs/usr/ports/databases/postgresql-plv8js/work/plv8-1.4.8'
*** Error code 1

Stop.
make: stopped in /usr/ports/databases/postgresql-plv8js
=>> Cleaning up wrkdir
===>  Cleaning for postgresql10-plv8js-1.4.8_6
build of databases/postgresql-plv8js | postgresql10-plv8js-1.4.8_6 ended at Mon Oct 18 09:41:51 CDT 2021
build time: 00:01:11
!!! build failure encountered !!!
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

I'll have a look as soon as
a) my package builder is done building my i386 repo and
b) I find some time

Always happy to fix something in ports, I'm pretty sure it can be fixed, but no guarantees I can do it ;-)
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 12,755
Messages: 39,353

So far I've gone in and modified the Makefile
Modify the PORTVERSION to 3.0.0 and remove PORTREVISION so it actually fetches the latest source code. Leave the compiler settings as is. Then see what happens. As this is a new version you probably need to run make makesum first, the port's distinfo needs to be updated for the new version. Then run make patch, this will apply the patches that are in the files/ directory. They're likely going to fail to apply (due to the changed sources), figure out which ones are still needed, remove those that don't apply anymore.
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

s:probably:certainly:

Doing that right now ;-)

(edit: I decided to start looking while my builder is still busy, so testing is pretty slow. Some day, I need a "super computer", hehe)
 
OP
F

freejetz888

New Member

Reaction score: 5
Messages: 18

Yeah, one dependency is lang/v8. This might take a while 😩
Yes, 3 hours to build on my machine. I do really appreciate your and SirDice's help.

On that, I've modified the Makefile to PORTVERSION= 3.0.0 and PORTREVISION= 1. I ran make makesum which finished successfully. I then ran make patch and removed all files that were failing, which was all of them. I then get the following error on build:

===> Building for postgresql10-plv8js-3.0.0_1
gmake[1]: Entering directory '/wrkdirs/usr/ports/databases/postgresql-plv8js/work/plv8-3.0.0'
mkdir -p build
cd build; git clone https://chromium.googlesource.com/chromium/tools/depot_tools.git
/bin/sh: git: not found
gmake[1]: *** [Makefile:55: build/depot_tools] Error 127
gmake[1]: Leaving directory '/wrkdirs/usr/ports/databases/postgresql-plv8js/work/plv8-3.0.0'
*** Error code 1

I installed git on this machine, reboot, and run again but receive the same error.

ETA: I can run git clone https://chromium.googlesource.com/chromium/tools/depot_tools.git just fine on its own.
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

The build tries to do a git clone? Oh my, this will be "fun" to fix. I WILL have a look, I can't promise anything, but in any case, don't expect a result too soon, it will probably take a few days…
I installed git on this machine, reboot, and run again but receive the same error.
Well, I guess you use poudriere? (and, you should!) -- then installing something locally on the machine won't have any effect. poudriere creates a jail for building that has nothing but what is actually listed as a dependency in the port itself :) – but a git clone during build is a no-go anyways, once the build starts, there should be no more "fetching". I guess this might be a PITA to port :-/
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 12,755
Messages: 39,353

PORTREVISION= 1
Should be 0 or removed. It's the first revision of the port with the 'new' version, so it starts counting at 0. Should you make a change to the port, without affection the version, then you should increase the PORTREVISION.
 

astyle

Daemon

Reaction score: 662
Messages: 1,482

There's a port that's been marked as "Broken" that I did manage to successfully compile: audio/lash. Relevant part to this thread - I told the Makefile to ignore the 'Broken' flag. After that, compiling went fine (The reason for the 'Broken' tag was a py27 dep, anyway). But fighting past source code errors (incorrectly specified APIs / variables) is a bit more than I have the patience for at this point.

Sometimes, I just look at what else is available in ports to solve my problem... But in this case, it looks like the problem is getting some attention from people who actually know what it takes to go from a simple compile error to getting the ports system to play nice with the software.
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

Should be 0 or removed.
Actually, removed. Explicitly setting it to 0 is frowned upon. This is only an option with complex stuff like e.g. master-slave-ports.
Should you make a change to the port, without affection the version, then you should increase the PORTREVISION.
Just for completeness: Only if the built result will differ. So e.g. for a simple build fix, you don't bump PORTREVISION.
But in this case, it looks like the problem is getting some attention from people who actually know what it takes to go from a simple compile error to getting the ports system to play nice with the software.
Well, I hope I know, at least I have some experience ;). It will take time until I can really start. Right now, my builder machine has a load of 61 (with 4 cores with hyperthreading). Uhm, yeah 😏
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

Here you go: PR 259262.

Not sure everything is correct (I can't test whether it works), but it's now building and packaging fine for me.

For testing, if you use git, just download the patch and apply it with git am. Otherwise, I assume the following should work: Make sure you have no local changes to the port, then go to the root of your ports tree and apply the patch with
fetch -o- "https://bz-attachments.freebsd.org/attachment.cgi?id=228826&action=diff&format=raw&headers=1" | patch -Np1

The build tries to do a git clone? Oh my, this will be "fun" to fix.
Luckily, this was only a minor problem. Upstream added a new Makefile that attempts to fetch and build all dependencies and links them statically :rolleyes: – nothing you'd ever want for a port or package. But the original Makefile doing just the build is still there, only renamed, so all it took to get out of this madness was a line MAKEFILE= Makefile.shared ;)
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

Ah damn, pointy-hat to me. I just discovered 3.0.0 seems to build fine with any postgresql version, so no need any more to restrict it to 9.6 :rolleyes:.

Added another patch to the PR, let's see ;)

freejetz888 if you use the default postgres version (currently 12) and want to use this port with it, as long as my additional fix isn't committed, try this little change to the port's Makefile after updating your ports tree:
Code:
-USES=        compiler:c++14-lang gmake perl5 pgsql:9.6-10
+USES=        compiler:c++14-lang gmake perl5 pgsql
 
OP
F

freejetz888

New Member

Reaction score: 5
Messages: 18

Amazing! Really appreciate your efforts Zirias! I just updated the ports tree and the package built successfully. I am still stuck using Postgres 10 due to the ERP application restrictions but I will keep the last trick in mind once I'm able to move past PG10. Also, looking at your PR 259262 it looks like someone also stepped up to be a maintainer. I will definitely look into getting better at this side of things and help maintain as well.

Again, thank you!
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 12,755
Messages: 39,353

I am still stuck using Postgres 10
Set DEFAULT_VERSIONS+= pgsql=10 in /etc/make.conf. Then everything you build from ports will be build against PostgreSQL 10. You can change the defaults when you build from ports. See /usr/ports/Mk/bsd.default-versions.mk.
 

Zirias

Son of Beastie

Reaction score: 1,702
Messages: 2,870

Also, looking at your PR 259262 it looks like someone also stepped up to be a maintainer.
Yep, I was surprised to see that ;) I'm maintaining a few ports myself, but I refused here because I think it's difficult to maintain something you don't use yourself…
I will definitely look into getting better at this side of things and help maintain as well.
That's great! One thing I really like about FreeBSD is that every user is welcome to contribute and make things better. Of course, it's a learning curve. But the existing ports framework is pretty nice and many things are easy to do (once you learned how, of course) ;)
 
OP
F

freejetz888

New Member

Reaction score: 5
Messages: 18

Yep, I have the default versions set for the Poudriere jails using the make.conf files. I've been using it for a few years so I'm pretty comfortable with this side of things. I would love to stay on the cutting edge but the applications we use seem to move at a snails pace when it comes to that.

Zirias, any advice or resources you can point me to that helped you become better at port/package maintenance? I know about the Porters Handbook, but I'm curious if any other resources are out there.
 
Top