Wireshark · Wireshark-dev: Re: [Wireshark-dev] Install failure at configure stage

[Joe Breher <linux@xxxxxxxxxxx>]

Ulf Lamping wrote:
> Joe Breher wrote:
>   
> > Jeff Morriss wrote:
> >   
> >     
> > > Sorry, my post was a question to the developers to help make it easier 
> > > for people in general.
> > >
> > > Anyway, for your case, just reconfigure Wireshark like this:
> > >
> > > ./configure --with-ssl
> > >
> > >   
> > >     
> > >       
> > This worked. Thanx.
> > A note in the user-guide-us.pdf might have staved off my inquiry.
> >   
> >     
> You might provide a detailed text and at which place to put this into 
> the guide.
>   
OK - I am working here from the supposedly current version of 
user-guide-us.pdf - at least it is the version that I downloaded about 
2.5 minutes ago from http://wireshark.org/docs/ .
I am on pdf page 32, doc page 18, section 2.4 (Building Wireshark from 
source under UNIX), numbered bullet 3. I am quoting herein the entire 
text of bullets 2, 3, and 4 for context - to wit:
"
2. Change directory to the Wireshark source directory.
3. Configure your source so it will build correctly for your version of 
UNIX. You can do this with the
following command:
./configure
If this step fails, you will have to rectify the problems and rerun 
configure. Troubleshooting hints
are provided in Section 2.6, “Troubleshooting during the install on Unix”.
4. Build the sources into a binary, with the make command. For example:
make
"

For those that have been following along, I encountered an error at 
bulleted item 3, where the ./configure failed due to an improper build 
option for UCD-SNMP. The fix was to do the configure step by the 
following invocation:
./configure --with-ssl .

I personally do not (yet) have the breadth of experience to know whether 
or not the default should be to invoke ./configure with the --with-ssl 
option. If so, then here in item 3 is where the fix to the install 
procedure should be made. Again assuming this is proper, the text here 
would change to be:
'
3. Configure your source so it will build correctly for your version of 
UNIX. You can do this with the
following command:
./configure --with-ssl
If this step fails, you will have to rectify the problems and rerun 
configure. Troubleshooting hints
are provided in Section 2.6, “Troubleshooting during the install on Unix”.
'

I rather suspect however, that there may be something about my system 
that makes it non-typical (comments?) in its need for a non-default 
build of UCD-SNMP. Changing our assumption to this being the case, the 
above suggested change is probably not appropriate. There is another 
section of the doc that deals with exceptions encountered during the 
build. We are now directed by bulleted item 3 to section 2.6, 
“Troubleshooting during the install on Unix”. We find this section on 
pdf page 35, doc page 21, quoted below in its entirety - to wit:
"
2.6. Troubleshooting during the install on Unix
A number of errors can occur during the installation process. Some hints 
on solving these are provided
here.
If the configure stage fails, you will need to find out why. You can 
check the file config.log in the
source directory to find out what failed. The last few lines of this 
file should help in determining the
problem.
The standard problems are that you do not have GTK+ on your system, or 
you do not have a recent
enough version of GTK+. The configure will also fail if you do not have 
libpcap (at least the required
include files) on your system.
Another common problem is for the final compile and link stage to 
terminate with a complaint of: Output
too long. This is likely to be caused by an antiquated sed (such as the 
one shipped with Solaris).
Since sed is used by the libtool script to construct the final link 
command, this leads to mysterious problems.
This can be resolved by downloading a recent version of sed from 
http://directory.fsf.org/GNU/
sed.html.
If you cannot determine what the problems are, send mail to the 
wireshark-dev mailing list explaining
your problem, and including the output from config.log and anything else 
you think is relevant, like
a trace of the make stage.
"

I would suggest that if, as Jeff Morriss suggests, this issue is 
encountered on a more-than-infrequent basis, the text of this section 
may be changed to read:
'
2.6. Troubleshooting during the install on Unix
A number of errors can occur during the installation process. Some hints 
on solving these are provided
here.
If the configure stage fails, you will need to find out why. You can 
check the file config.log in the
source directory to find out what failed. The last few lines of this 
file should help in determining the
problem.
The standard problems are that you do not have GTK+ on your system, or 
you do not have a recent
enough version of GTK+. The configure will also fail if you do not have 
libpcap (at least the required
include files) on your system.
The configure stage may sometimes fail due to an incompatible 
configuration of ucd-snmp. If the last few lines of the configure stage 
fail with a console indication of an error involving 'UCD SNMP', the 
error may be skirted by passing an option to the configuration process. 
This can often be addressed by configuring with the invocation
./configure --with-ssl.
Another common problem is for the final compile and link stage to 
terminate with a complaint of: Output
too long. This is likely to be caused by an antiquated sed (such as the 
one shipped with Solaris).
Since sed is used by the libtool script to construct the final link 
command, this leads to mysterious problems.
This can be resolved by downloading a recent version of sed from 
http://directory.fsf.org/GNU/
sed.html.
If you cannot determine what the problems are, send mail to the 
wireshark-dev mailing list explaining
your problem, and including the output from config.log and anything else 
you think is relevant, like
a trace of the make stage.
'


The astute observer will not that I have inserted the fourth paragraph. 
It reflects an ignorant noob's view of what may or may not be going on 
under the hood. I imagine there is likely a better way to address this 
issue, which I shall learn all about in time. All I know, is that had 
this paragraph been in the user-guide-us.pdf, I would not yet have met 
all you fine contributors on wireshark-dev.