An interesting statement is "!: not found"

Can you run with on the first line #!/bin/sh -xv instead of #!/bin/sh and attach the entire output to this bug report.

Seems like 

if ! test -d "generated_src/doxygen"; then
  mkdir -p generated_src/doxygen
fi
if test "$f_wizard" = YES; then
  if ! test -d "generated_src/doxywizard"; then
    mkdir -p generated_src/doxywizard
  fi
fi

is indeed not understood by your shell. Can you check if you replace it with

mkdir -p generated_src/doxygen
if test "$f_wizard" = YES; then
  mkdir -p generated_src/doxywizard
fi

There is no harm in trying to create an existing directory with mkdir -p.

Yes, this fixes the issue. There is now another issue left at the end of configure:

  Generating generated_src/doxygen/lang_cfg.h...
Traceback (most recent call last):

+ Trace 233503

  Generating generated_src/doxygen/settings.h...
  Finished
        [configure-modulated] complete for oxygen.

However, compilation seems to start fine.

Strange as the statement is "protected" against "underflow" of argv.

The file reads:
import sys

if (len(sys.argv) > 0):
    if (sys.argv[1] == "ENONLY"):
        print "#define ENGLISH_ONLY"
    else:
        for x in xrange(1, len(sys.argv)):
            print "#define LANG_%s"%(sys.argv[x])

Can You change this in:
import sys

print (sys.version)
print len(sys.argv)
print sys.argv
if (len(sys.argv) > 0):
    if (sys.argv[1] == "ENONLY"):
        print "#define ENGLISH_ONLY"
    else:
        for x in xrange(1, len(sys.argv)):
            print "#define LANG_%s"%(sys.argv[x])

and send the result.

This is probably due to the length indexing starting by 0:

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > /opt/csw/bin/python src/lang_cfg.py
2.6.9 (unknown, Oct 31 2013, 16:41:40) [C]
1
['src/lang_cfg.py']
Traceback (most recent call last):

+ Trace 233504

zsh: 12171 exit 1     /opt/csw/bin/python src/lang_cfg.py
dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > /opt/csw/bin/python src/lang_cfg.py abc
2.6.9 (unknown, Oct 31 2013, 16:41:40) [C]
2
['src/lang_cfg.py', 'abc']
#define LANG_abc

The ./configure script should look after it that lang_cfg.py always gets arguments. In case no languages are specified the entire list of languages should follow.
Can you also do the test with ./configure? and send the content of generated_src/doxygen/lang_cfg.h


In configure the variables f_lang and f_ulang are used, would be good to have their content as well just before calling src/lang_cfg.py
(also the output with #!/bin/sh -xv would do)

The full configure with 'set -x' is available at
  http://buildfarm.opencsw.org/~dam/doxygen-1.8.7-configure.out
The reason is that f_langs is empty because sed on Solaris is not GNU sed:

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > ls -1 src/translator_??.h | sed -e 's%src/translator_%%g' | sed -e 's/\.h//' | tr '\012' ',' ;echo
am,ar,br,ca,cn,cz,de,dk,en,eo,es,fa,fi,fr,gr,hr,hu,id,it,je,jp,ke,kr,lt,lv,mk,nl,no,pl,pt,ro,ru,sc,si,sk,sr,sv,tr,tw,ua,vi,za,
dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > ls -1 src/translator_??.h | sed -e 's%src/translator_%%g' | sed -e 's/\.h//' | tr '\012' ',' | sed -e 's/,$//';echo

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > (ls -1 src/translator_??.h | sed -e 's%src/translator_%%g' | sed -e 's/\.h//' | tr '\012' ',' ;echo)| sed -e 's/,$//';echo
am,ar,br,ca,cn,cz,de,dk,en,eo,es,fa,fi,fr,gr,hr,hu,id,it,je,jp,ke,kr,lt,lv,mk,nl,no,pl,pt,ro,ru,sc,si,sk,sr,sv,tr,tw,ua,vi,za

It looks like Solaris sed does not like unterminated strings as input:

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > printf "abc" | sed -e 's/b/d/';echo

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 >

Created attachment 274971 [details]
Test script to be run on solarsi for sed problem

As you mentioned the sed on Solaris is not GNO sed, is it possible to obtain a version of GNU sed for Solaris ?

It is for me not clear which of the sed commands causes the problem, so can you run the attached script?

Is it possible to point out what the right syntax might be for the troublesome sed command? If we cannot find a solution we have to think about implementing this line in something else (e.g. awk).

Sure:

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > ./solaris_tst.sh 
======= step 1 ==========
src/translator_am.h src/translator_ar.h src/translator_br.h src/translator_ca.h src/translator_cn.h src/translator_cz.h src/translator_de.h src/translator_dk.h src/translator_en.h src/translator_eo.h src/translator_es.h src/translator_fa.h src/translator_fi.h src/translator_fr.h src/translator_gr.h src/translator_hr.h src/translator_hu.h src/translator_id.h src/translator_it.h src/translator_je.h src/translator_jp.h src/translator_ke.h src/translator_kr.h src/translator_lt.h src/translator_lv.h src/translator_mk.h src/translator_nl.h src/translator_no.h src/translator_pl.h src/translator_pt.h src/translator_ro.h src/translator_ru.h src/translator_sc.h src/translator_si.h src/translator_sk.h src/translator_sr.h src/translator_sv.h src/translator_tr.h src/translator_tw.h src/translator_ua.h src/translator_vi.h src/translator_za.h
======= step 2 ==========
am.h ar.h br.h ca.h cn.h cz.h de.h dk.h en.h eo.h es.h fa.h fi.h fr.h gr.h hr.h hu.h id.h it.h je.h jp.h ke.h kr.h lt.h lv.h mk.h nl.h no.h pl.h pt.h ro.h ru.h sc.h si.h sk.h sr.h sv.h tr.h tw.h ua.h vi.h za.h
======= step 3 ==========
am ar br ca cn cz de dk en eo es fa fi fr gr hr hu id it je jp ke kr lt lv mk nl no pl pt ro ru sc si sk sr sv tr tw ua vi za
======= step 4 ==========
am,ar,br,ca,cn,cz,de,dk,en,eo,es,fa,fi,fr,gr,hr,hu,id,it,je,jp,ke,kr,lt,lv,mk,nl,no,pl,pt,ro,ru,sc,si,sk,sr,sv,tr,tw,ua,vi,za,
======= step 5 ========== 
 
========================= 

A normal Solaris 10 does not ship GNU sed, so relying on that would be bad. It looks like that
sed works on reading full lines terminated by a newline. I think the last line without newline is
silently ignored. The XPG4 sed shipped with Solaris gives a hint on that:

dam@unstable10s [unstable10s]:/home/dam > printf "a,b,c" | /usr/xpg4/bin/sed -e 's/b/d/';echo
sed: Missing newline at end of file standard input.
a,d,c

I am not sure if this behavior of Solaris sed can be modified. 

Unfortunately my awk is not that good, in Perl I would write it like this:

dam@unstable10s [unstable10s]:/home/dam/mgar/pkg/doxygen/trunk/work/solaris10-sparc/build-isa-sparcv8plus/doxygen-1.8.7 > ls -1 src/translator_??.h | sed -e 's%src/translator_%%g' | sed -e 's/\.h//' | perl  -e 'local $/;print join(",",split(/\n/,<>)),"\n"' 
am,ar,br,ca,cn,cz,de,dk,en,eo,es,fa,fi,fr,gr,hr,hu,id,it,je,jp,ke,kr,lt,lv,mk,nl,no,pl,pt,ro,ru,sc,si,sk,sr,sv,tr,tw,ua,vi,za

I just had a chat on IRC on #sed and User emg suggested to _not parse ls_(tm) as in
  http://mywiki.wooledge.org/ParsingLs

Instead he suggested the posix-conformant construct
  unset comma; for f in src/translator_??.h; do f=${f%.h}; printf %s%s "$comma" "${f#src/translator_}"; comma=, done

Created attachment 275069 [details]
Updated configure script

I had a look at it and I think the solution can be quite simple, just remove the last sed in the f_langs assignment. The last comma is removed (at least where I test it when translating f_langs to f_ulangs. If this is not the case under Solaris (can be checked by printing f_ulangs), I just have to adjust the lang_cfg.py slightly (ignoring empty arguments), but I prefer when this is not necessary.

I've attached an updated configure (including the changes for the if ( ! test  problem.

Please try it, the last line in the generated/doxygen/lang_cfg.h should be
#define LANG_ZA
if it works

Yes, that line is the last line. configure looks good now.

Thank you for all the information.
I've just pushed a patch to github (pull request 161) including all changes.

This bug fix fixes, in a general way, a number of issues encountered on Solaris

- configure
  - ! (exclamation mark) not recognized properly, but with mkdir -p no test is necessary (fix from Dimitri)
  - sed problem regarding removing last , (comma) in language list. Is automatically replaced in with a space when translating language names to upper case.
- lang_cfg.py
  - corrected check on number of arguments

Pull request integrated in git

I have now set up continuous integration for doxygen building on Solaris:
  https://buildfarm.opencsw.org/buildbot/waterfall?builder=doxygen-solaris10-i386&builder=doxygen-solaris10-sparc&reload=60
Feel free to link to the page for reference. I can also set up notification on build failures.

This bug was previously marked ASSIGNED, which means it should be fixed in
doxygen version 1.8.8. Please verify if this is indeed the case. Reopen the
bug if you think it is not fixed and please include any additional information 
that you think can be relevant (preferrably in the form of a self-contained example).