gdb-apple @1824: build failure due to texinfo5

[cooljeanius (Eric Gallager)]

The error looks like this:

./elf.texi:1: warning: @section missing argument
bfd.texinfo:335: unknown command `colophon'
bfd.texinfo:346: unknown command `cygnus'
./bfdt.texi:313: raising the section level of @subsubsection which is too low
./init.texi:5: raising the section level of @subsubsection which is too low
./archive.texi:54: raising the section level of @subsubsection which is too low
./format.texi:27: raising the section level of @subsubsection which is too low
./reloc.texi:495: raising the section level of @subsubsection which is too low
./core.texi:8: raising the section level of @subsubsection which is too low
./opncls.texi:5: raising the section level of @subsubsection which is too low
./libbfd.texi:10: raising the section level of @subsubsection which is too low
./cache.texi:14: raising the section level of @subsubsection which is too low
./elf.texi:13: raising the section level of @subsubsection which is too low
make[2]: *** [bfd.info] Error 1
make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_devel_gdb-apple/gdb-apple/work/gdb-1824/src-x86_64/bfd/doc'
make[1]: *** [info-recursive] Error 1
make[1]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_devel_gdb-apple/gdb-apple/work/gdb-1824/src-x86_64/bfd'
make: *** [all-bfd] Error 2
make: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_devel_gdb-apple/gdb-apple/work/gdb-1824/src-x86_64'
Command failed:  cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_devel_gdb-apple/gdb-apple/work/gdb-1824/src-x86_64" && /usr/bin/make -w all MAKEINFO="/opt/local/bin/makeinfo" LEXLIB="/opt/local/lib/libfl.a" 
Exit code: 2
Error: org.macports.build for port gdb-apple returned: command execution failed
DEBUG: Error code: NONE
DEBUG: Backtrace: command execution failed

Whenever I ran into this error when building my fork of gdb-apple manually, I would always work around it by making sure that a copy of texinfo4 appeared first in my path... I guess the solution here would be to use the system makeinfo (i.e. /usr/bin/makeinfo) here instead of the MacPorts one, as that is still version 4, at least on my machine:

Local-Admins-MacBook-Pro:~ ericgallager$ /usr/bin/makeinfo --version
makeinfo (GNU texinfo) 4.8

Copyright (C) 2004 Free Software Foundation, Inc.
There is NO warranty.  You may redistribute this software
under the terms of the GNU General Public License.
For more information about these matters, see the files named COPYING.

[cooljeanius (Eric Gallager)]

Oops, my logfile for the original error got overwritten by the one from when it built successfully after I modified the Portfile...

[larryv (Lawrence Velázquez)]

Cc Me!

[ntraft@…]

Hey Eric, could you post instructions for what changes are necessary to the Portfile?

The confusing thing here is that even though it output an error, I still have a gdb-apple binary in my bin/ folder, and it seems to work just fine... makeinfo also looks like it's there... are these prebuilt binaries? Why am I seeing this error?

[ntraft@…]

Cc Me!

[cooljeanius (Eric Gallager)]

Replying to ntraft@…:

Hey Eric, could you post instructions for what changes are necessary to the Portfile?

Well I have made a bunch of changes to the portfile in general, so I am not really sure which ones were the necessary ones... I suppose I can try to walk through them though...

The first hunk is mostly just whitespace changes, so we can skip over it (I like my lines wrapped a little tighter than normal):

Portfile

@@ -6 +6 @@
 
 name            gdb-apple
 version         1824
+revision        1
 categories      devel
 license         GPL-2+
 maintainers     jeremyhu openmaintainer
-description     GDB: The GNU Project Debugger with Apple Modifications (based on 6.3.50)
+description     GDB: The GNU Project Debugger with Apple Modifications \
+                (based on 6.3.50)
 
 long_description \
-GDB, the GNU Project debugger, allows you to see what is going on 'inside' \
-another program while it executes -- or what another program was doing at the \
-moment it crashed.  GDB can do four main kinds of things (plus other things \
-in support of these) to help you catch bugs in the act: \
-    a) start your program, specifying anything that might affect its behavior, \
-    b) make your program stop on specified conditions, \
-    c) examine what has happened, when your program has stopped, \
-    d) change things in your program, so you can experiment with correcting \
-       the effects of one bug and go on to learn about another. \
-The program being debugged can be written in Ada, C, C++, Objective-C, \
-Pascal (and many other languages). Those programs might be executing on \
-the same machine as GDB (native) or on another machine (remote). GDB \
-can run on most popular UNIX and Microsoft Windows variants.
+GDB, the GNU Project debugger, allows you to see what is going on \
+'inside' another program while it executes -- or what another program was \
+doing at the moment it crashed. GDB can do four main kinds of things \
+(plus other things in support of these) to help you catch bugs in action: \
+\n\
+    a) start your program, specifying anything that might affect its \
+       behavior, \n\
+    b) make your program stop on specified conditions, \n\
+    c) examine what has happened, when your program has stopped, \n\
+    d) change things in your program, so you can experiment with \
+       correcting the effects of 1 bug and go on to learn about another.\n\
+\n\
+The program being debugged can be written in Ada, C, C++, Objective-C, or \
+Pascal, as well as many other languages. Those programs might be \
+executing on the same machine as GDB (native) or on another machine \
+(remote). GDB can run on most popular UNIX and Microsoft Windows variants \
+(this version is primarily for OS X though).
 
 homepage        http://opensource.apple.com/source/gdb
 platforms       darwin

In the next hunk I removed the dependency on texinfo and replaced it with a dependency on gsed, because there is some tricky sed usage that I was trying to silence a warning from... gsed is not strictly necessary though.
I also added some library dependencies that otool -L said were linked against, and that nm -m confirmed that symbols were used from; but those are not relevant to the makeinfo parts though.
I also added a comment explaining why worksrcdir has to be changed.
Also all the "-append"s are because I always use "-append" by default, just in case, unless I specifically need to override something. Those additions can be ignored though:

Portfile

@@ -39 +45 @@
 checksums       rmd160  90290c3950dd06c1691a477b9ee302c00b681ed1 \
                 sha256  d25b056a826015e1e0ea338b9b8f03e40dce3ff4696faf95df259ae1a42698da
 
-depends_build   port:gettext port:zlib port:flex port:texinfo
-
-worksrcdir      ${worksrcdir}/src
+depends_build-append \
+                port:gettext \
+                port:zlib \
+                port:flex \
+                port:gsed
+# Removed texinfo because we now force the use of the system makeinfo
+
+# port-depcheck.sh reports these linkages:
+depends_lib-append \
+                port:libiconv \
+                port:ncurses \
+                port:sqlite3
+
+# The configure script is one level deeper; the top-level Makefile is
+# hand-written and full of Apple-internal-specific things:
+worksrcdir      ${worksrcdir}/src
 
 post-patch {
     # https://trac.macports.org/ticket/37790

In this next hunk I tried some additional configure arguments. None of them led to any actual changes though, so they can probably be ignored:

Portfile

@@ -51 +70 @@
         ${worksrcpath}/bfd/configure.host
 }
 
-configure.args \
+configure.args-append \
     --infodir=${prefix}/share/info \
     --mandir=${prefix}/share/man \
     --with-docdir=${prefix}/share/doc \
     --program-suffix=-apple \
-    --disable-werror
+    --disable-werror \
+    --with-mpfr=${prefix} \
+    --with-gmp=${prefix} \
+    --enable-libada \
+    --enable-libssp
+

In this next hunk I tried setting a bunch of environment variables.
The GSED one forces the use of the gsed port that I added a build dependency on above.
I just realized that the values to MAKEINFO are different in the configure and build environments, but that is okay, because either one works, at least on my machine. If we want to be consistent, though, it might make more sense to stick with the one with missing in it, as that would be safer.
All the variables with "MSG" somewhere in them, and the other variables around there, are to try to force the usage of the gettext port from MacPorts. They are not strictly necessary though.
A lot of the variables with "RC_" in them, and the other variables around there, are to try to simulate being inside a darwinbuild environment, which was part of my attempts to get the project to build from the top-level Makefile instead of from the one that we currently use in src/. I never managed to get that to work though, so those variables can probably be ignored as well.
Clearing the value of READELF_PROG was because I had been experiencing a build failure with it, and this was an attempt to try to fix that. Although that might have just been copied from my work on the Portfile for the FSF version of gdb, so I am not sure if it is strictly necessary here as well...
(Note that this hunk of the diff is not actually a separate hunk; I just split it off from the previous one so that I could fit these comments in here)

Portfile

@@ +83 @@
+
+configure.env-append \
+    SED=${prefix}/bin/gsed \
+    MSGFMT=${prefix}/bin/msgfmt \
+    GMSGFMT=${prefix}/bin/msgfmt \
+    XGETTEXT=${prefix}/bin/xgettext \
+    MSGMERGE=${prefix}/bin/msgmerge \
+    MAKEINFO="${worksrcpath}/missing makeinfo" \
+    INSTALLED_PRODUCT_ASIDES=YES \
+    RC_JASPER=YES \
+    RC_NONARCH_CFLAGS="${configure.optflags}" \
+    RC_XBS=YES \
+    SEPARATE_STRIP=YES \
+    UNAME_RELEASE="${os.version}" \
+    HOST_ARCHITECTURE="${configure.build_arch}"
+
+platform macosx {
+    configure.args-append \
+        --enable-carbon-framework \
+        --enable-debug-symbols-framework
+    configure.env-append \
+        RC_OS=macos \
+        RC_TARGET_CONFIG=MacOSX
+    build.env-append \
+        RC_OS=macos \
+        RC_TARGET_CONFIG=MacOSX
+}
+
+platform darwin {
+    configure.env-append \
+        RC_PRIVATE=/private \
+        UNAME_SYSNAME=Darwin \
+        NEXT_ROOT=""
+    build.env-append \
+        RC_PRIVATE=/private \
+        UNAME_SYSNAME=Darwin \
+        NEXT_ROOT=""
+}
+
+build.args-append \
+    MAKEINFO="/usr/bin/makeinfo" \
+    LEXLIB="${prefix}/lib/libfl.a" \
+    READELF_PROG=""
+
+build.env-append \
+    MAKEINFO="/usr/bin/makeinfo" \
+    LEXLIB="${prefix}/lib/libfl.a" \
+    SED=${prefix}/bin/gsed \
+    MSGFMT=${prefix}/bin/msgfmt \
+    GMSGFMT=${prefix}/bin/msgfmt \
+    XGETTEXT=${prefix}/bin/xgettext \
+    MSGMERGE=${prefix}/bin/msgmerge \
+    INSTALLED_PRODUCT_ASIDES=YES \
+    RC_JASPER=YES \
+    RC_XBS=YES \
+    SEPARATE_STRIP=YES \
+    UNAME_RELEASE="${os.version}" \
+    HOST_ARCHITECTURE="${build_arch}" \
+    READELF_PROG=""
 
-build.args \
-    MAKEINFO="${prefix}/bin/makeinfo" \
+destroot.args-append \
+    MAKEINFO="/usr/bin/makeinfo" \
     LEXLIB="${prefix}/lib/libfl.a"
 
-destroot.args \
-    MAKEINFO="${prefix}/bin/makeinfo" \
+destroot.env-append \
+    MAKEINFO="/usr/bin/makeinfo" \
    LEXLIB="${prefix}/lib/libfl.a"
 
 post-destroot {

Finally, in this last hunk, instead of deleting stuff, I solved some conflicts with other ports by renaming the files instead of deleting them. I also had destroot install some additional files that I considered to be useful. This part is all taking place after the place where the makeinfo failure had occurred though, so these next changes are not actually strictly relevant to this issue:

Portfile

@@ -74 +156 @@
         move ${destroot}${prefix}/share/info/${info} ${destroot}${prefix}/share/info/apple-${info}
     }
 
-    eval delete [glob ${destroot}${prefix}/lib/*.{,l}a]
-    eval delete [glob ${destroot}${prefix}/bin/{addr2line,ar,c*filt,nm,objcopy,objdump,ranlib,readelf,size,strings,strip}-apple]
+    # avoid conflicts with other gdb/binutils installations from MacPorts:
+    xinstall -d ${destroot}${prefix}/lib/${name}
+    eval move [glob ${destroot}${prefix}/lib/*.{,l}a] ${destroot}${prefix}/lib/${name}
+    eval reinplace "s|${prefix}/lib|${prefix}/lib/${name}|g" [glob ${destroot}${prefix}/lib/${name}/*.la]
+    xinstall -d ${destroot}${prefix}/include/${name}
+    eval move [glob ${destroot}${prefix}/include/*.h] ${destroot}${prefix}/include/${name}
+    xinstall -d ${destroot}${prefix}/share/info/${name}
+    eval move [glob ${destroot}${prefix}/share/info/{annotate,bfd,stabs}.info] ${destroot}${prefix}/share/info/${name}
+    foreach msgs [glob ${destroot}${prefix}/share/locale/*/LC_MESSAGES/*.mo] {
+        if [file isfile ${msgs}] {
+            move ${msgs} [file dirname ${msgs}]/${name}-[file tail ${msgs}]
+        }
+    }
+    xinstall -d ${destroot}${prefix}/share/${name}
+    eval xinstall -m 644 [glob ${worksrcpath}/../conf/*.gdb] ${destroot}${prefix}/share/${name}
+    xinstall -d ${destroot}${prefix}/etc/${name}
+    xinstall ${worksrcpath}/../gdb.conf ${destroot}${prefix}/etc/${name}
+    xinstall -d ${destroot}${prefix}/share/doc/${name}
+    xinstall ${worksrcpath}/../doc/refcard.pdf ${destroot}${prefix}/share/doc/${name}
     eval delete [glob ${destroot}${prefix}/*darwin*]
-    eval delete [glob ${destroot}${prefix}/include/*.h]
-    eval delete [glob ${destroot}${prefix}/share/info/{annotate,bfd,stabs}.info]
-    eval delete [glob ${destroot}${prefix}/share/locale/*/LC_MESSAGES/{bfd,binutils,opcodes}.mo]
 }
 
 if {${os.platform} eq "darwin" && ${os.major} >= 12} {

So, to summarize, the changes that fixed the texinfo issue for me here were to append MAKEINFO="${worksrcpath}/missing makeinfo" to the configure environment, and to change the MAKEINFO="${prefix}/bin/makeinfo" entry in the build and destroot environment to be MAKEINFO="/usr/bin/makeinfo" instead, and to remove the dependency on texinfo. Those ones are the relevant ones here; the rest can be ignored.

Replying to ntraft@…:

The confusing thing here is that even though it output an error, I still have a gdb-apple binary in my bin/ folder, and it seems to work just fine... makeinfo also looks like it's there... are these prebuilt binaries? Why am I seeing this error?

Which bin/ folder? The one in your prefix? It could just be a previous version. And makeinfo already being in the bin/ folder in your prefix is what was causing this issue in the first place. If you meant that a new gdb-apple binary still gets built, though, it could just be parallelism in make causing the libraries that gdb-apple needs to get built before the documentation...

[jeremyhu (Jeremy Huddleston Sequoia)]

can you please just attach the patch?

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #45515.

[rphoover@…]

Patch for Yosemite

[rphoover@…]

Replying to ryandesign@…:

Has duplicate #45515.

The file Portfile-gdb-apple.diff is a proposed patch for gdb-apple to work with Yosemite. This patch allows the program to build and initial testing with Lazarus (Free Pascal) indicates gdb-apple works.

[jeremyhu (Jeremy Huddleston Sequoia)]

Why do you need gdb? I'd rather just remove the project instead as it is not nearly as helpful as lldb these days.

[jeremyhu (Jeremy Huddleston Sequoia)]

Submitted in r127586, but I still want to know what your use case is.

[rphoover@…]

Replying to jeremyhu@…:

Why do you need gdb? I'd rather just remove the project instead as it is not nearly as helpful as lldb these days.

Lazarus does not support lldb yet. When it does, then there will be no need for this project.

[cooljeanius (Eric Gallager)]

Replying to jeremyhu@…:

Why do you need gdb? I'd rather just remove the project instead as it is not nearly as helpful as lldb these days.

For me at least I prefer gdb for political reasons (it's "free-er" than lldb is), and because I'm more familiar with it, and because it's not written in C++ like lldb is. Plus also lldb doesn't have a port for it yet (as seen in #45251 for example). Now I can see maybe removing the gdb-apple port if the upstream FSF gdb project merged all of Apple's changes into their trunk, but given that I've been working for over a year on that (in my own fork) with little progress, I doubt it will happen any time soon...