wireshark: update to 2.0.1

[ryandesign (Ryan Carsten Schmidt)]

​wireshark 2.0.1 is out; you may want to update the wireshark[-devel] port to this version. Note that with this version comes a major ​change from gtk to qt.

[smith.kennedy+macports@…]

Cc Me!

[smith.kennedy+macports@…]

Please update this port as well to include fix from #40620 because the same problem with DT_UNKNOWN exists in the wireshark-devel port as well.

[hsivank@…]

This is a rewrite of the original Wireshark Portfile.
It is possible to switch between Qt4, Qt5, Gtk2, Gtk3 backend.
The configure step is done by cmake. (As Qt, it will be the future of Wireshark...)
For now I use svn to fetch source. (There are some problems with official 1.11.0 tar : bugs, files missing ...)

[ryandesign (Ryan Carsten Schmidt)]

• Could you provide a unified diff instead of a complete Portfile?
• The revision should drop to 0 when the version is increased.
• You're using the cmake portgroup which automatically adds a cmake build dependency so you don't need to manually add one (just make sure you append to depends_build instead of overwriting it).
• Unless you separately need subversion at build time, you should not add a subversion build dependency; MacPorts automatically adds a subversion fetch dependency when fetch.type is svn.
• Variants whose names begin with negative words like "no_" have been deprecated for years and we should work on removing them from ports; we should not add new variants with such names (like the "no_gui" variant you added here). If the user wants no gui, they would deselect any default variants that add guis.

[hsivank@…]

Thanks for your comments.
What about the qt5 variant ? Macports does not support Qt5 yet.

[ryandesign (Ryan Carsten Schmidt)]

Right, we can't have a qt5 variant here until #37331 is resolved.

[hsivank@…]

ok i will remove it

[hsivank@…]

Portfile_wireshark-devel-2.diff doesn't have qt5 variant

[ryandesign (Ryan Carsten Schmidt)]

Looks like I was mistaken about the no_gui variant; it has already existed in the wireshark-devel port for years. But it's confusing then that it's not in the wireshark port too. I wish the wireshark and wireshark-devel ports were maintained by the same maintainer(s) and did not differ so wildly from one another. A port and its -devel counterpart should typically only differ in the version number and any necessary changes to make that version build properly.

You don't need to set CMAKE_INSTALL_NAME_DIR; the cmake portgroup already sets it for you. You just need to append to configure.args rather than overwriting it. You also still need to append to, rather than overwriting, depends_build.

There are a lot of variants in this port. Why has a zlib variant been added, for example? Under what circumstances would a user want to disable zlib support? Variants are typically added when an optional dependency takes a lot of time or disk space to install, or when the user must choose amongst multiple options.

[hsivank@…]

Currently Wireshark is evolving a lot, so I think it is the good moment to remove old variant .. if not you will have *no_gui* variant for 10 years again in macports ;-)
If i remove CMAKE_INSTALL_NAME_DIR, RPATH is removed during destroot phase. (osx 10.7)
Variants reflect option provided by wireshark : ​http://anonsvn.wireshark.org/viewvc/trunk/CMakeOptions.txt?revision=52513&view=markup

Yes there is a big diff between wireshark and wireshark-devel. But cmake is the way to go. When qtshark will be released as stable, The Portfile will be ready for Wireshark's Portfile.

[ryandesign (Ryan Carsten Schmidt)]

Replying to hsivank@…:

Currently Wireshark is evolving a lot, so I think it is the good moment to remove old variant .. if not you will have *no_gui* variant for 10 years again in macports ;-)

Yes. wireshark-devel also still has a "no_ssl" variant which should be restructured to be an "ssl" variant. wireshark still has variants no_adns, no_geoip, no_gnutls, no_ipv6, no_libgcrypt, no_libsmi, no_lua, no_rtp, no_ssl, no_x11; fixes to restructure these should be brought over to the wireshark port from the wireshark-devel port.

If i remove CMAKE_INSTALL_NAME_DIR, RPATH is removed during destroot phase. (osx 10.7)

I don't see why it would matter whether the portfile adds the option or the portgroup adds the option. Either way, the option should get added.

Variants reflect option provided by wireshark : ​http://anonsvn.wireshark.org/viewvc/trunk/CMakeOptions.txt?revision=52513&view=markup

A portfile usually should not offer variants for every option supported by the build system. Rather, the maintainer should select the options that make sense and program them into the portfile, and offer a very small number of variants if necessary to support unusual options or ones with heavy dependencies.

Yes there is a big diff between wireshark and wireshark-devel. But cmake is the way to go. When qtshark will be released as stable, The Portfile will be ready for Wireshark's Portfile.

I understand that wireshark 1.11 has switched to cmake, so that difference between wireshark and wireshark-devel is expected. The other differences, including differing variants and differing whitespace, are not. In this case, the variants and whitespace of the wireshark-devel port are more conformant with current MacPorts best practices, so the wireshark port should adopt them.

[hsivank@…]

Replying to ryandesign@…:

Replying to hsivank@…:

If i remove CMAKE_INSTALL_NAME_DIR, RPATH is removed during destroot phase. (osx 10.7)

I don't see why it would matter whether the portfile adds the option or the portgroup adds the option. Either way, the option should get added.

I fix the RPATH problem : configure.args was used instead of configure.args-append.

A portfile usually should not offer variants for every option supported by the build system. Rather, the maintainer should select the options that make sense and program them into the portfile, and offer a very small number of variants if necessary to support unusual options or ones with heavy dependencies.

I propose to support only qt, gtk2, gtk3 and pythons variants.

[mojca (Mojca Miklavec)]

Cc Me!

[mojca (Mojca Miklavec)]

Are there any news about this update? There is version 1.11.2 out already and there should be no problem to fetch the sources from tar.bz2 (if there are, they also have git access available now in addition to svn).

Personally I would vote for providing the qt interface only. I don't see any particular need for gtk, it will only cause headaches and one will have to fix problems related to it without any benefit. It only makes sense to support gtk if Qt isn't working properly yet. I'm not saying that the port shouldn't provide it, only that it's not necessary to have it. You can keep support for GTK if you find it useful.

And yes, I also suggest to simply turn all the relevant features on without a special variant for that unless that feature isn't working, causes problems or requires too many dependencies.

Can you please submit a new Portfile (or diff) for 1.11.2 and taking the latest comments into account? (Ryan explicitly asked for a diff. This usually makes a lot of sense, but in this particular case where basically everything other than description is changed, plain Portfile works much better, at least for me. I'm not advocating using one or another, only explaining that in cases like this one the diffs are hardly more informative.)

I agree with the proposal to leave the old wireshark port intact (= not to worry about it) and simply rename this port once the version 1.12.0 gets released.

[hsivank@…]

Wireshark 1.11.2 does not build on osx with cmake and Qt gui is not ready. This path (Portfile_wireshark-devel-1.11.2.diff) update current Portfile with minimum modification. (it still use autotools, Gtk …)

[mojca (Mojca Miklavec)]

One question (out of many): ./configure --help says

  --with-broken-python[=DIR]

Doesn't that mean that the option name also needs to be adapted? I get:

configure: WARNING: unrecognized options: --without-pcre, --without-python

Let's forget for a moment that cmake-based installation isn't ready yet (I didn't even try it). What about providing the option to enable Qt? I built it with

    --without-gtk2 \
    --without-gtk3 \
    --with-qt \
    --with-osx-integration

and it compiled fine at least. I downloaded their Qt binary for mac and it also seems to work just fine. I didn't play much with the port from MacPorts though.

Anyway, I can apply your patch as is (without Qt if you didn't test it yet), but it would make sense to figure out what to do with the python switch (whether it needs to be renamed or not).

[hsivank@…]

python support is broken ​https://anonsvn.wireshark.org/viewvc?revision=49138&view=revision, so we can remove it …
about pcre : it seems that it is not used anymore
I can build gtk2 + Qt4 (macports) interface, but it fails with gtk3 => i will add a variant for Qt4

[mf2k (Frank Schima)]

I applied your patch and ran lint. Here are the results:

$ port lint --nitpick 
--->  Verifying Portfile for wireshark-devel
Warning: Variant gnutls does not have a description
Warning: Variant libgcrypt does not have a description
Warning: Variant ipv6 does not have a description
Warning: Variant lua does not have a description
Warning: Variant no_ssl does not have a description
Warning: Variant libsmi does not have a description
Warning: Dependency port:gtk2 specified multiple times in depends_lib
--->  0 errors and 7 warnings found.

The last one needs to be fixed. Indeed gtk2 is always a dependency and the default gtk2 adds it again. Adding descriptions to all variants would be very helpful too. Finally, the no_ssl variant name needs to be changed to ssl and thus do the opposite. Is there a good reason for it? Maybe it should simply be removed.

[mojca (Mojca Miklavec)]

I'm sorry, I failed to notice the attachment (trac doesn't send emails when attachments are added). Just a question: how does the software behave when both GTK and Qt are being used? (I would imagine that the two options need to conflict each other, but I might be wrong. What happens when the user double-clicks WireShark.app and both GTK and Qt are enabled? I'm sorry for asking without testing myself.)

[mojca (Mojca Miklavec)]

Based on ​this commit I wonder if wireshark actually works better with Qt 5 on Mac.

[hsivank@…]

Replying to mojca@…:

how does the software behave when both GTK and Qt are being used? (I would imagine that the two options need to conflict each other, but I might be wrong. What happens when the user double-clicks WireShark.app and both GTK and Qt are enabled? I'm sorry for asking without testing myself.)

Currently we don't build bundle. Two executables are created inside /opt/local/bin. wireshark with Gtk and qtshark with Qt.

[hsivank@…]

Replying to macsforever2000@…:

I applied your patch and ran lint. Here are the results: The last one needs to be fixed. Indeed gtk2 is always a dependency and the default gtk2 adds it again. Adding descriptions to all variants would be very helpful too. Finally, the no_ssl variant name needs to be changed to ssl and thus do the opposite. Is there a good reason for it? Maybe it should simply be removed.

Thanks for your remarks. I will try to make a proper patch quickly

[hsivank@…]

Replying to mojca@…:

Based on ​this commit I wonder if wireshark actually works better with Qt 5 on Mac.

Yes, official Wireshark binaries are provided with Qt5

[ryandesign (Ryan Carsten Schmidt)]

qt5 is in MacPorts now.

[hadrielk@…]

Wireshark's development release is up to 1.99.7 now (and will be 1.99.8 tomorrow), and there's a separate Ticket #48363 for that, so you might want to close this ticket or dup it to #48363.

[hsivank@…]

Updated my previous Portfile with cmake support and clean variant description

~$ port variants wireshark-devel
wireshark-devel has the variants:
[+]adns: use adns library for async. dns resolution
[+]cares: Build with c-ares support
   debug: Enable debug binaries
[+]geoip: Build with GeoIP support
[+]gnutls: Build with GNU TLS support
   gtk2: Use GTK3 instead of GTK2 to build wireshark
     * conflicts with gtk3 no_gui
   gtk3: Build Wireshark (GTK+ UI)
     * conflicts with gtk2 no_gui
   kerberos5: Build with Kerberos support
[+]libgcrypt: Build with GNU crypto support
[+]libsmi: Build with libsmi snmp support
[+]lua: Build with Lua dissector support
   no_gui: Disable wireshark and wireshark_gtk
     * conflicts with gtk2 gtk3 qt4 qt5
[+]portaudio: Build with PortAudio support
   python25: use python25 for the experimental python interface
     * conflicts with python26 python27
   python26: use python26 for the experimental python interface
     * conflicts with python25 python27
   python27: use python27 for the experimental python interface
     * conflicts with python25 python26
   qt4: Use Qt5 instead of Qt4 to build wireshark
     * conflicts with no_gui qt5
[+]qt5: Build Wireshark
     * conflicts with no_gui qt4
   universal: Build for multiple architectures
[+]zlib: Build with zlib compression support

[mdeaudelin (Mathieu Deaudelin-Lemay)]

Replying to hsivank@…:

Updated my previous Portfile with cmake support and clean variant description [...]

Some comments about this patch:

Variants:

• Python support has been removed, so those variants are setting configure options which are being ignored.
• There is no use setting both cares and adns as default variants, only one of them will be applied.
• Setting qt5 as a default variant makes it impossible to build with qt4, since they conflict. An "if" is needed:
  

  if {![variant_isset qt]} { 
  default_variants    +qt5
  }
  

• Setting the gtk2 or gtk3 variants builds a qt5 binary AND a gtk binary. Is that intentional?
• The wireshark port offers variants to enable or disable ipv6 and ssl, kerberos5 is enabled by default and variants gtk2 and portaudio are named x11 and rtp respectively. It may be good to offer compatible variants so that this port can be used as a drop-in replacement once wireshark 1.99 becomes the stable build.
• There is an error at line 182, for variant geoip, where you wrote kerberos5 instead.

Dependencies:

• Variant gtk3 is missing a dependency for port:adwaita-icon-theme, which I had added in my patch.
• "depends_lib port:openssl" on line 40 is reverting the patch made in changeset 139229: ​https://trac.macports.org/changeset/139229/trunk/dports/net/wireshark-devel/Portfile

Changes:

• These calls:
  "eval xinstall -m 644 [glob ${worksrcpath}/epan/*.h] ${destroot}${prefix}/include/wireshark/epan/"
  should look like this:
  "xinstall -m 644 {*}[glob ${worksrcpath}/epan/*.h] ${destroot}${prefix}/include/wireshark/epan/" (see #48363, comment #1).

Application Launcher:

• As I noted in #48363, comment #9, this patch does not create launchers in the Application folder for the gtk2 and gtk3 variants.
• For qt4/qt5, the application launcher created starts with a lowercase w, while the convention is to have it uppercase.

[mdeaudelin (Mathieu Deaudelin-Lemay)]

Cc Me!

[hsivank@…]

Replying to macports@…:

• Python support has been removed, so those variants are setting configure options which are being ignored.

fixed

• There is no use setting both cares and adns as default variants, only one of them will be applied.

fixed

• Setting qt5 as a default variant makes it impossible to build with qt4, since they conflict. An "if" is needed:
  

you can use

port install wireshark +qt4 -qt5

• Setting the gtk2 or gtk3 variants builds a qt5 binary AND a gtk binary. Is that intentional?

Well, it will be problematic if i provide bundle for each variant ... so it will not be possible anymore

• The wireshark port offers variants to enable or disable ipv6 and ssl,

This options are not provided by cmake (I ask myself : Is there a reason to disable ipv6 or ssl ?)

kerberos5 is enabled by default

Kerberos5 breaks build for now, so it is not enabled by default

variants gtk2 and portaudio are named x11 and rtp respectively. It may be good to offer compatible variants so that this port can be used as a drop-in replacement once wireshark 1.99 becomes the stable build.

i agree, maybe we can make *fake* x11 and rtp variants to support legacy ?

• There is an error at line 182, for variant geoip, where you wrote kerberos5 instead.

fixed

• Variant gtk3 is missing a dependency for port:adwaita-icon-theme, which I had added in my patch..

fixed

• "depends_lib port:openssl" on line 40 is reverting the patch made in changeset 139229: ​https://trac.macports.org/changeset/139229/trunk/dports/net/wireshark-devel/Portfile

fixed

• These calls:
  "eval xinstall -m 644 [glob ${worksrcpath}/epan/*.h] ${destroot}${prefix}/include/wireshark/epan/"
  should look like this:
  "xinstall -m 644 {*}[glob ${worksrcpath}/epan/*.h] ${destroot}${prefix}/include/wireshark/epan/" (see #48363, comment #1).

fixed

• As I noted in #48363, comment #9, this patch does not create launchers in the Application folder for the gtk2 and gtk3 variants.

to be tested (only tested gtk2) ...

• For qt4/qt5, the application launcher created starts with a lowercase w, while the convention is to have it uppercase.

fixed

[mojca (Mojca Miklavec)]

• checksums: you should use rmd160 and sha256. You may remove md5 and sha1.
• depends_build: use depends_build-append instead and remove dependency on cmake.
• configure.args: use configure.args-append unless you have to overwrite all settings set by the cmake group.
• you should not use default_variants +qt5; you should first check that the user did not set +qt4 (ask if you need advice on coding that, there are plenty examples in other ports, but basically you just need to remove the variant and put it in the "if {![variant_isset qt4] && ![variant_isset qt5]} {" block); or rather, you should also check that the user did not specify +gtk3 before enabling qt5 (if you plan to keep support for gtk3 and if you want the variants to conflict)
• enable things like lua and zlib by default

[hsivank@…]

Replying to mojca@…:

• checksums: you should use rmd160 and sha256. You may remove md5 and sha1.

I would prefer *official* wireshark hash (md5,sha1,ripemd160) : ​https://www.wireshark.org/download/src/all-versions/SIGNATURES-1.99.8.txt

• depends_build: use depends_build-append instead and remove dependency on cmake.

fixed

• configure.args: use configure.args-append unless you have to overwrite all settings set by the cmake group.

fixed

• you should not use default_variants +qt5; you should first check that the user did not set +qt4 (ask if you need advice on coding that, there are plenty examples in other ports, but basically you just need to remove the variant and put it in the "if {![variant_isset qt4] && ![variant_isset qt5]} {" block); or rather, you should also check that the user did not specify +gtk3 before enabling qt5 (if you plan to keep support for gtk3 and if you want the variants to conflict)

default_variants +portaudio +zlib +lua +libsmi +gnutls +libgcrypt +cares +geoip

if {![variant_isset qt4] && ![variant_isset qt5] && ![variant_isset gtk2] && ![variant_isset gtk3] && ![variant_isset no_gui]} {
	default_variants +qt5
}

Is it correct ?

• enable things like lua and zlib by default

Maybe i'm missing something but zlib and lua are already enabled by default

default_variants ... +zlib +lua  ...

[mdeaudelin (Mathieu Deaudelin-Lemay)]

Replying to hsivank@…:

This options are not provided by cmake (I ask myself : Is there a reason to disable ipv6 or ssl ?)

Indeed, I don't see a good reason to disable them either. I'm just comparing to the stable wireshark port, but I'd bet very few people turn those off.

Kerberos5 breaks build for now, so it is not enabled by default

Is that specific to the cmake build? I had no trouble building with kerberos enabled without cmake.

i agree, maybe we can make *fake* x11 and rtp variants to support legacy ?

Good point, especially since your variants have more descriptive names. Aliases can be created if and when they are needed (e.g. variant x11 requires gtk2).

Replying to hsivank@…:

Replying to mojca@…:

• checksums: you should use rmd160 and sha256. You may remove md5 and sha1.

I would prefer *official* wireshark hash (md5,sha1,ripemd160) : ​https://www.wireshark.org/download/src/all-versions/SIGNATURES-1.99.8.txt

I had made the same call in my proposed patch. Using the same checksums as upstream makes it easier to compare and verify.

default_variants +portaudio +zlib +lua +libsmi +gnutls +libgcrypt +cares +geoip

if {![variant_isset qt4] && ![variant_isset qt5] && ![variant_isset gtk2] && ![variant_isset gtk3] && ![variant_isset no_gui]} {
	default_variants +qt5
}

Is it correct ?

That's pretty much what the current wireshark port uses, so that sounds right IMO.

[mojca (Mojca Miklavec)]

Replying to hsivank@…:

Replying to mojca@…:

• checksums: you should use rmd160 and sha256. You may remove md5 and sha1.

I would prefer *official* wireshark hash (md5,sha1,ripemd160) : ​https://www.wireshark.org/download/src/all-versions/SIGNATURES-1.99.8.txt

What if you just add sha256 then?

default_variants +portaudio +zlib +lua +libsmi +gnutls +libgcrypt +cares +geoip

if {![variant_isset qt4] && ![variant_isset qt5] && ![variant_isset gtk2] && ![variant_isset gtk3] && ![variant_isset no_gui]} {
	default_variants +qt5
}

Is it correct ?

Yes, I think so. (Unless you invert no_gui into gui of course.)

Just for fun I tried what it meant switching to +gui. The weird fact is that with the usage of

default_variants +portaudio +zlib +lua +libsmi +gnutls +libgcrypt +cares +geoip +gui

if {[variant_isset gui]} {
    if {![variant_isset qt4] && ![variant_isset qt5] && ![variant_isset gtk2] && ![variant_isset gtk3]} {
    	default_variants +qt5
    }
}

variant qt5 requires gui conflicts qt4 gtk2 gtk3 description {Build Wireshark} {...}

MacPorts should complain when using

port -v configure wireshark-devel -gui +qt5

But it doesn't. I'm wondering if this is a bug in MacPorts.

• enable things like lua and zlib by default

Maybe i'm missing something but zlib and lua are already enabled by default

default_variants ... +zlib +lua  ...

Sorry, I meant unconditionally enable them (not even giving the users any choice). Both dependencies are very small and I'm not sure if it brings any added value to the port to allow turning those variants off. But that was just a suggestion that you don't need to follow.

[hsivank@…]

Replying to macports@…:

Kerberos5 breaks build for now, so it is not enabled by default

Is that specific to the cmake build? I had no trouble building with kerberos enabled without cmake.

CMake module embded in Wireshark to detect kerberos is not really correct (it miss k5crypto)
Patching against this module is a waste of time. I propose add link flag at configure time :

variant kerberos5 description {Build with Kerberos support} {
        configure.args-append   -DENABLE_KERBEROS=ON \
                                -DCMAKE_SHARED_LINKER_FLAGS="-lk5crypto"
        depends_lib-append      port:kerberos5
}

i agree, maybe we can make *fake* x11 and rtp variants to support legacy ?

Good point, especially since your variants have more descriptive names. Aliases can be created if and when they are needed (e.g. variant x11 requires gtk2).

It will be great. Could a MacPort guru confirm it is a good idea ?

[hsivank@…]

Replying to mojca@…:

What if you just add sha256 then?

fixed

Sorry, I meant unconditionally enable them (not even giving the users any choice). Both dependencies are very small and I'm not sure if it brings any added value to the port to allow turning those variants off. But that was just a suggestion that you don't need to follow.

I would like to respect options provided in wireshark/CMakeOptions.txt

[mdeaudelin (Mathieu Deaudelin-Lemay)]

Oh and I forgot: You may want to add the following at the end of the Portfile:

livecheck.type  regex 
livecheck.url   ${homepage}download.html 
livecheck.regex "Development Release \\((\\d+(?:\\.\\d+)*)" 

This will allow the ​livecheck feature to work.

[hsivank@…]

Replying to macports@…:

Oh and I forgot: You may want to add the following at the end of the Portfile:

livecheck.type  regex 
livecheck.url   ${homepage}download.html 
livecheck.regex "Development Release \\((\\d+(?:\\.\\d+)*)" 

This will allow the ​livecheck feature to work.

Fixed

[hsivank@…]

It seems that we have something acceptable.
Maybe it's time to provide diff against current wireshark-devel Portfile ...

[hsivank@…]

fix trailingspace issue

[mike-savory]

I just tested the latest "Portfile-cmake-1.99.5.8​" on a fully updated (10.11.1) ElCapitan powerbook with Xcode 7.1.1. I had to change qt5-mac to qt5 (in 2 locations) to get rid of these errors

Error: qt5-mac has been replaced by qt5; please install that instead.

Error: org.macports.configure for port qt5-mac returned: obsolete port

And then updated version to 2.0.0 and updated the checksums

rmd160 79b1badcb34ce5d4d6b8c43de7d0ed2c7c301527

sha256 90026c761a85701d7783c7e2eaa4c1de247dfbadbd53221df355f121e42691dd

It compiles, and runs from the console (/opt/local/bin/wireshark) but fails to run by double clicking /Applications/MacPorts/Wireshark.app (I assume this may be a ElCap protection issue).
Good news for Ticket #48835

[mike-savory]

Cc Me!

[mike-savory]

Digging deeper, the error from trying to run /Applications/MacPorts/Wireshark.app is

Dyld Error Message:
  Library not loaded: @rpath/QtPrintSupport.framework/Versions/5/QtPrintSupport
  Referenced from: /Applications/MacPorts/Wireshark.app/Contents/MacOS/Wireshark
  Reason: image not found

But it looks to be installed by macports

-- Qt5PrintSupport includes: /opt/local/libexec/qt5/lib/QtPrintSupport.framework;/opt/local/libexec/qt5/lib/QtPrintSupport.framework/Headers;/opt/local/libexec/qt5/lib/QtWidgets.framework;/opt/local/libexec/qt5/lib/QtWidgets.framework/Headers;/opt/local/libexec/qt5/lib/QtGui.framework;/opt/local/libexec/qt5/lib/QtGui.framework/Headers;/opt/local/libexec/qt5/lib/QtCore.framework;/opt/local/libexec/qt5/lib/QtCore.framework/Headers;/opt/local/libexec/qt5/.//mkspecs/macx-clang;/System/Library/Frameworks/OpenGL.framework/Headers
-- Qt5PrintSupport libs: Qt5::PrintSupport
-- Qt5PrintSupport definitions: -DQT_PRINTSUPPORT_LIB;-DQT_WIDGETS_LIB;-DQT_GUI_LIB;-DQT_CORE_LIB

[mojca (Mojca Miklavec)]

Cc Me!

[mojca (Mojca Miklavec)]

I didn't test it, but something like this should work:

[sudo] install_name_tool -change \
    @rpath/QtPrintSupport.framework/Versions/5/QtPrintSupport \
    /opt/local/libexec/qt5/lib/QtPrintSupport.framework/Versions/5/QtPrintSupport \
    /Applications/MacPorts/Wireshark.app/Contents/MacOS/Wireshark

[mojca (Mojca Miklavec)]

Cc Me!

[mike-savory]

Thank Mojca, yes that gets me one step closer, now it errors on

Dyld Error Message:
  Library not loaded: @rpath/QtMultimedia.framework/Versions/5/QtMultimedia
  Referenced from: /Applications/MacPorts/Wireshark.app/Contents/MacOS/Wireshark
  Reason: image not found

So I guess it is a systemic issue.... I guess rinse and repeat (but not tonight, finished here for the day) as a workaround.

Mike

[mojca (Mojca Miklavec)]

I know you explained what you did to the Portfile, but it would nevertheless be helpful if you would upload the Portfile for version 2.0.0 with all your changes. Given that it's a final release it might even be wireshark, not just wireshark-devel.

Someone should look into the building scripts. It might be a trivial patch or a trivial build setting to fix this. Just run

otool -L /Applications/MacPorts/Wireshark.app/Contents/MacOS/Wireshark

and you'll get a list of all libraries that Wireshark expects.

[ghosthound]

Replying to mojca@…:

I know you explained what you did to the Portfile, but it would nevertheless be helpful if you would upload the Portfile for version 2.0.0 with all your changes. Given that it's a final release it might even be wireshark, not just wireshark-devel.

Note that the plan is to make the current 'wireshark' port into 'wireshark1' and then have 'wireshark' move to wireshark-2.0.0 (see ​https://trac.macports.org/ticket/48835#comment:43).

[mike-savory]

Personally I would vote for continuing the existing wireshark port as a GTK default build of the 1.x stream and adding a new port for wireshark2 that defaults to a qt build. Managing the transition to a qt based build should be a deliberate choice, not the default upgrade path for existing wireshark users. That way we can make a clean break from the previous variants.

[mike-savory]

Edited port file for wireshark 2.0 with corrected checksums

[mike-savory]

It now launches from the /Applications/MacPorts/Wireshark.app, The otool showed which libraries to change.

The "install_name_tool" required sudo and needed to be run for the following libraries

QtPrintSupport QtMultimedia QtSvg QtMacExtras QtNetwork QtWidgets QtGui QtCore

[hsivank@…]

Maybe it's time to test ENABLE_APPLICATION_BUNDLE option! Not sure if it will break build again ...

[mojca (Mojca Miklavec)]

See also #49837.

[hsivank@…]

looking at #49837, it is a better idea to fix rpath in Qt rather than using ENABLE_APPLICATION_BUNDLE ...

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

Replying to hsivank@…:

looking at #49837, it is a better idea to fix rpath in Qt rather than using ENABLE_APPLICATION_BUNDLE ...

rpath was removed from Qt in r143113.

[mojca (Mojca Miklavec)]

I'm attaching a Portfile for wireshark 2.0.1 using CMake. It is heavily based on msavory's template, but with a number of changes and not all the variants have been tested yet.

I kept post-destroot as added by msavory, but I commented it out as I don't yet see the added value of it (Wireshark can do it automatically; not sure if also for GTK). I didn't test GTK3 yet, but it works nicely with quartz (if I just replace the version in the old configure-based installation starting from wireshark 1).

I would suggest to retire wireshark-devel for now (replaced_by wireshark). If there is a need to keep Wireshark 1, a "svn cp wireshark wireshark1" from current and minor corrections should do (or major ones if there is need to keep the ports more similar to each other).

See also the port abandonement ticket in #50243.

[mojca (Mojca Miklavec)]

Update wireshark to version 2.0.1 (CMake-based installation)

[smith.kennedy+macports@…]

As a longtime user of Wireshark via MacPorts, and a frequent user of Wireshark, I would prefer that a "Wireshark 1" port be preserved, for at least the short term. I don't know if doing this would be impractical or awkward from a port maintenance point of view, but I could imagine a new "wireshark2" port being created, let users beat on that, and once it is deemed "stable", that could be phased in as the new "wireshark" port, and let the existing 1.x port become the "wireshark1" port.

[hsivank@…]

Replying to mojca@…:

Looks good !

I kept post-destroot as added by msavory, but I commented it out as I don't yet see the added value of it (Wireshark can do it automatically; not sure if also for GTK).

post-destroot was added before cause ENABLE_APPLICATION_BUNDLE was *bugged*

I would suggest to retire wireshark-devel for now (replaced_by wireshark). If there is a need to keep Wireshark 1, a "svn cp wireshark wireshark1" from current and minor corrections should do (or major ones if there is need to keep the ports more similar to each other).

+1

See also the port abandonement ticket in #50243.

[ghosthound]

minor update: I've taken the portfile from mojca, made a few adjustments and renamed it to 'wireshark2'. Testing is next.

[mike-savory]

I have built both 2.0.0 and 2.0.1 without problems over the last month with qt5 and cmake, they should be the defaults in my opinion! I can test (I have 10.11.3, 10.11.4b Server and 10.10.5 available)

[mike-savory]

The post-destroot was only a temporary fix as qt5 was evolving, it should be removed now.

Here is my last 2.0.1 build info (Thought I had lua included before, let me try again...)

Compiled (64-bit) with Qt 5.5.1, with libpcap, without POSIX capabilities, with libz 1.2.8, with GLib 2.46.2, with SMI 0.5.0, with c-ares 1.10.0, without Lua, with GnuTLS 3.3.19, with Gcrypt 1.6.4, with MIT Kerberos, with GeoIP, with QtMultimedia, without AirPcap. Running on Mac OS X 10.11.3, build 15D21 (Darwin 15.3.0), with locale C, with libpcap version 1.7.4, with libz 1.2.8, with GnuTLS 3.3.20, with Gcrypt 1.6.4. Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz (with SSE4.2) Built using clang 4.2.1 Compatible Apple LLVM 7.0.2 (clang-700.1.81).

[ghosthound]

(hopefully) resolved in #145542. Added a new port, net/wireshark2 based on the attached Portfiles, thanks for the initial work. Please test this port, I've built it with qt5 (default) and it seems to work fine, other UI builds have not been tested. Added as wireshark2 to reduce the impact on current wireshark (1.x) users.

[mojca (Mojca Miklavec)]

Link to the commit: r145542

[Schamschula (Marius Schamschula)]

Quite honestly, I'm not a fan of QT5, given how unstable it's build process seems to be.

However, this is not my issue here. I tried building with

sudo port install wireshark2 +gtk2 +lua -qt5

which most closely resembles my existing install of wireshark 2.0.1, see #48835.

With wireshark2 I get the following error:

:info:build /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/wslua/lua_bitop.c:85:2: error: "Unknown number type, check LUA_NUMBER_* in luaconf.h"
:info:build #error "Unknown number type, check LUA_NUMBER_* in luaconf.h"
:info:build  ^
:info:build 1 error generated.
:info:build cd /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan && /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang  -DG_DISABLE_DEPRECATED -DG_DISABLE_SINGLE_INCLUDES -DWS_BUILD_DLL -Depan_EXPORTS -isystem /opt/local/include -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/tools/lemon -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/wiretap -isystem /opt/local/include/glib-2.0 -isystem /opt/local/lib/glib-2.0/include -isystem /opt/local/include/gtk-2.0 -isystem /opt/local/include/freetype2 -isystem /opt/local/include/atk-1.0 -isystem /opt/local/include/gdk-pixbuf-2.0 -isystem /opt/local/include/cairo -isystem /opt/local/include/pango-1.0 -isystem /opt/local/lib/gtk-2.0/include -isystem /opt/local/include/lua-5.2 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/dfilter -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/dissectors -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/ftypes -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/wslua -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/nghttp2  -Wall -W -Wextra -Wendif-labels -Wpointer-arith -Warray-bounds -Wformat-security -fwrapv -fno-strict-overflow -Wvla -Waddress -Wattributes -Wdiv-by-zero -Wignored-qualifiers -Wpragmas -Wno-overlength-strings -Wwrite-strings -Wno-long-long -Wheader-guard -Wunused-const-variable -Qunused-arguments -Wc++-compat -Wdeclaration-after-statement -Wshadow -Wno-pointer-sign -Wold-style-definition -Wstrict-prototypes -Wshorten-64-to-32 -pipe -Os  -fvisibility=hidden -DNDEBUG -arch x86_64 -mmacosx-version-min=10.10 -fPIC   -o CMakeFiles/epan.dir/wslua/wslua_byte_array.c.o   -c /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/wslua/wslua_byte_array.c
:info:build make[2]: *** [epan/CMakeFiles/epan.dir/wslua/lua_bitop.c.o] Error 1
:info:build make[2]: *** Waiting for unfinished jobs....
:info:build [ 24%] Building C object epan/CMakeFiles/epan.dir/wslua/wslua_capture_info.c.o
:info:build cd /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan && /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang  -DG_DISABLE_DEPRECATED -DG_DISABLE_SINGLE_INCLUDES -DWS_BUILD_DLL -Depan_EXPORTS -isystem /opt/local/include -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/tools/lemon -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/wiretap -isystem /opt/local/include/glib-2.0 -isystem /opt/local/lib/glib-2.0/include -isystem /opt/local/include/gtk-2.0 -isystem /opt/local/include/freetype2 -isystem /opt/local/include/atk-1.0 -isystem /opt/local/include/gdk-pixbuf-2.0 -isystem /opt/local/include/cairo -isystem /opt/local/include/pango-1.0 -isystem /opt/local/lib/gtk-2.0/include -isystem /opt/local/include/lua-5.2 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/dfilter -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/dissectors -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/ftypes -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/wslua -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/nghttp2  -Wall -W -Wextra -Wendif-labels -Wpointer-arith -Warray-bounds -Wformat-security -fwrapv -fno-strict-overflow -Wvla -Waddress -Wattributes -Wdiv-by-zero -Wignored-qualifiers -Wpragmas -Wno-overlength-strings -Wwrite-strings -Wno-long-long -Wheader-guard -Wunused-const-variable -Qunused-arguments -Wc++-compat -Wdeclaration-after-statement -Wshadow -Wno-pointer-sign -Wold-style-definition -Wstrict-prototypes -Wshorten-64-to-32 -pipe -Os  -fvisibility=hidden -DNDEBUG -arch x86_64 -mmacosx-version-min=10.10 -fPIC   -o CMakeFiles/epan.dir/wslua/wslua_capture_info.c.o   -c /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1/epan/wslua/wslua_capture_info.c
:info:build make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1'
:info:build make[1]: *** [epan/CMakeFiles/epan.dir/all] Error 2
:info:build make[1]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1'
:info:build make: *** [all] Error 2
:info:build make: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1'
:info:build Command failed:  cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_net_wireshark2/wireshark2/work/wireshark-2.0.1" && /usr/bin/make -j8 -w all VERBOSE=ON 
:info:build Exit code: 2

All this means is that we're using the wrong lua headers. Using classic configure this is a non-issue. However, as you are now using cmake, I can't help you here…

[Schamschula (Marius Schamschula)]

Cc Me!

[mojca (Mojca Miklavec)]

The configuration looks fine:

-- Checking for one of the modules 'lua5.2;lua-5.2;lua52;lua5.1;lua-5.1;lua51;lua5.0;lua-5.0;lua50'
-- Found LUA: /opt/local/lib/lua-5.2/liblua-5.2.dylib (found version "502") 
-- LUA FOUND
-- LUA includes: /opt/local/include/lua-5.2
-- LUA libs: /opt/local/lib/lua-5.2/liblua-5.2.dylib

but the problem is in the order of arguments:

/path/to/wireshark2/work/wireshark-2.0.1/epan && /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang  -DG_DISABLE_DEPRECATED -DG_DISABLE_SINGLE_INCLUDES -DQT_CORE_LIB -DQT_GUI_LIB -DQT_NO_DEBUG -DWS_BUILD_DLL -Depan_EXPORTS -isystem /opt/local/include -I/path/to/wireshark2/work/wireshark-2.0.1 -I/path/to/wireshark2/work/wireshark-2.0.1/epan -I/path/to/wireshark2/work/wireshark-2.0.1/tools/lemon -I/path/to/wireshark2/work/wireshark-2.0.1/wiretap -isystem /opt/local/include/glib-2.0 -isystem /opt/local/lib/glib-2.0/include -isystem /opt/local/include/lua-5.2 -isystem /opt/local/libexec/qt4/include -isystem /opt/local/libexec/qt4/include/QtGui -isystem /opt/local/libexec/qt4/include/QtCore -I/opt/local/libexec/qt4/include/QtMultimedia -I/path/to/wireshark2/work/wireshark-2.0.1/epan/dfilter -I/path/to/wireshark2/work/wireshark-2.0.1/epan/dissectors -I/path/to/wireshark2/work/wireshark-2.0.1/epan/ftypes -I/path/to/wireshark2/work/wireshark-2.0.1/epan/wslua -I/path/to/wireshark2/work/wireshark-2.0.1/epan/nghttp2  -Wall -W -Wextra -Wendif-labels -Wpointer-arith -Warray-bounds -Wformat-security -fwrapv -fno-strict-overflow -fno-delete-null-pointer-checks -Wvla -Waddress -Wattributes -Wdiv-by-zero -Wignored-qualifiers -Wno-overlength-strings -Wwrite-strings -Wno-long-long -Qunused-arguments -Wc++-compat -Wdeclaration-after-statement -Wshadow -Wno-pointer-sign -Wold-style-definition -Wstrict-prototypes -Wshorten-64-to-32 -pipe -Os  -fvisibility=hidden -DNDEBUG -arch x86_64 -mmacosx-version-min=10.7 -fPIC   -o CMakeFiles/epan.dir/wslua/lua_bitop.c.o   -c /path/to/wireshark2/work/wireshark-2.0.1/epan/wslua/lua_bitop.c

That is '-isystem /opt/local/include' comes before '-isystem /opt/local/include/lua-5.2'.

[mojca (Mojca Miklavec)]

But please open a new ticket for that.