Recent changes to patches

Header file includes, namespace, gettext fix

pinder — Wed, 06 Apr 2005 12:52:08 -0000

Patch is againts CVS. Fixes include:

Linking againts libintl for gettext that the new GTK+
version use.

Some std:: namespaces have been fixed. This previously
did not allow complete compilation.

Includes have been converted to the new C++ standard of
<filename> rather than <filename.h>.

Roopinder.

Fix CXXFLAGS in Makefile

Steven Augart — Fri, 01 Mar 2002 22:20:43 -0000

Makefile-12.patch
Submitted by Steven Augart <saugart at users dot
sourceforge dot net>

Apply this in the root Althea source directory with:
patch -p1 < Makefile-clean-target-11.patch

This patch fixes some minor glitches where I'd confused
CPPFLAGS with CXXFLAGS. These would not have affected
compilation,
but do make the Makefile more consistent.

Added -fimplicit-templates to CXXFLAGS. This helps
call attention to
Althea's template instantiation model.

make clean should clean up dependencies

Steven Augart — Thu, 28 Feb 2002 07:15:50 -0000

This one-line patch adds the dependencies files (*.d) to
the Makefile's `clean' target.

Automatic Makefile dependencies

Steven Augart — Tue, 26 Feb 2002 01:46:05 -0000

Submitted by Steven Augart <saugart at users dot
sourceforge dot net>

Apply this in the root Althea source directory with:
patch -p1 <
automatic-Makefile-dependencies-10.patch

This patch solves a problem I encountered during
development
where out-of-date-dependencies led to an inconsistent
build and
a subsequent weird Althea crash.

Details:

*The Makefile automatically generates and updates
dependencies. It
does this by creating a batch of files suffixed .d, one
for each
source file. Got rid of the old and inadequate
hardcoded
dependencies.

*Now use standard names (e.g., CXXFLAGS instead of
CFLAGS) which
cuts out about 2K from the makefile, despite the new
targets.

* The necessary -ftemplate-depth-30 flag is no longer
misleadingly stuck
into DEBUGFLAGS

*Makefile now has configuration options at the start of
the Makefile.

*The Makefile now uses the GNU project's recommended
names for
various directories (e.g., $(bindir) instead of
$(DESTDIR}).
This helps for future compatibility with Autoconf.

Fix constructors.

Steven Augart — Sat, 16 Feb 2002 04:56:49 -0000

Patch: fix-constructors-8a.patch

This patch fixes bugs in various constructors. It also
defines an
explicit constructor and an explicit virtual destructor
for every
class.

This text looks prettier if you read it under Emacs
with M-x outline-mode.

*All of the constructors and destructors are now
defined in the
appropriate <class>.cpp file, instead of being inline
in the <class>.h
file. This reduces the size of the executable;
otherwise they get
redundantly instantiated in every source file they're
used in.

*All classes now have virtual destructors. This is
needed so that
destructors on polymorphic subclassed objects will work
properly
even when they're being referred to by a pointer to
objects of an
ancestral type.

*Bugs fixed in various constructors:
**class Althea: Needed a constructor to zero-initialize
some members.
This never showed up because the only instances of
objects of the
Althea class are both global variables, and with
globals, their memory
starts off automatically zeroed at program load time.
The result of *that* is that if someone ever *were* to
instantiate an object
of class Althea on the stack or via the memory manager,
the bug would
only appear then.

*Some classes had constructors that would have worked
in Java,
but sneakily fail in C++. I fixed them.

For example, the no-arg constructor:
Folder::Folder() {
Folder("New Folder");
}

did not initialize the new folder object's folder name.
Instead, it
invoked the constructor Folder(string), which then in
turn instantiated a
new Folder with its title set to "New Folder", and then
immediately
destroyed that new Folder!

The classes Server and Folder had this bug.

*Old Folder constructor did not initialize server_ptr.
Lurking bug. Fixed.

*strings are initialized to "" by default
so no need to do it again in a constructor. I've
excised such
initializations from the constructor code.

*Redundant code consolidated.
Some classes (such as Folder or Server) with more than
one constructor
now have each constructor call a single central
initialize() function.
This removed code duplication.

*All classes now have constructors and destructors.
Some of these are empty in the patches I'm submitting
here. They are important in my own work, though. In
*my* working tree, they
initialize some #ifndef NDEBUG internal consistency
checks (assertions
and the like) that I've already used in my own Althea
sources to
catch bad-pointer bugs quickly.

I do not plan to send patches for those consistency
checks unless one
of the core developers actually cares about it and
understands why
this is a good thing to have. The idea is well
discussed in Kernighan
and Pike's ``The Practice of Programming'' and in Steve
McConnell's
``Writing Solid Code''

Makefile: Better TAGS output for C++

Steven Augart — Fri, 15 Feb 2002 23:08:16 -0000

This one-line change to the Makefile makes the TAGS
output more useful.

Specifically, it tells etags that Althea's .h files are
C++ source, not C source. etags also now adds
information about member functions to the TAGS file.

Fix crash in on_check_mail_menu_activate

Steven Augart — Tue, 12 Feb 2002 21:05:26 -0000

Brief summary: Modifies on_check_mail_menu_activate()
in callbacks.cpp:

This describes modifications to the function
on_check_mail_menu_activate(). It fixes a crash that
happens if we
have already failed to read a folder list from the
server.

*Problems corrected:

**We never checked whether the_server->get_Folder()
succeeded.
(If we fail to read the folder list from the server,
then that
call won't succeed, and havoc will result -- I should
know; this
happened to me!).

**gAlthea.get_Current_Server() was being called after
we'd
already initialized the variable the_server with its
contents. This in turn deceived the programmer into
writing some redundant code to handle the pointer
Ťthe_serverť being NULL. This redundant code was never
executed, since that check had already been performed.

**We called the_server->get_Folder() before we had a
reservation for
the server.

*Enhancement
When there is no server available (because none has
been set up yet),
on_check_mail_menu_activate will now give the better
error message:
"Go to Althea ==> Setup... and define a server."
instead of the confusing:
"Select a server first."

*Problems not corrected (yet?):
**The server reservation (6 lines) is code repeated in
a number of places;
should be consolidated into a method of the Server
class (something
like reserve_server_blocking().

**There's a synchronization problem with the server
reservation code.
Only way to fix this is to go over to using a proper
Posix 1003.1c mutex.

**Other places need checks for whether a current server
exists.
I made on_expunge_menu_activate() dump core.

A couple of other functions do use assertions to
confirm that there's
a current server. A step in the right direction, but
obviously a user
action should never trigger an assertion.

It would be good to have this code modularized and put
it there. The
right thing to do is have an
Althea::get_current_server_or_gripe()
method.

*Future Enhancement:
**When <Print> is clicked, check for whether there's a
current message
before popping up the dialogue box, not after.

Misc Bug fixes (thank you, G++!)

Steven Augart — Tue, 12 Feb 2002 07:44:23 -0000

Summary: This patch contains fixes for the causes of
various warnings that were flagged by aggressive use of
the g++ warning options. At least some of these
warnings pointed out cases in the code that would have
caused crashes or undesired behavior.

Also includes additional conversions from using
integer error codes
to using the Ťenum errorsť error codes. Revised the
error code
system to use the predicates isSuccess() and
isFailure().

Unpatcking instructions included in the patch.

Long Description (this is identical to the contents of
the .msg-2 files):

MIME.cpp: Converted more code to use enum errors.
Added comments.
Commented out unused code.
Revised decode64() to be more robust: (a) added
assertions.
(b) added comments
(c) moved variables from function-scope into the
innermost
appropriate block.

MIME.h: Match changes in MIME.cpp

callbacks.cpp: Added error checking to the call to
decode64.
Improved error messages by calls to the new
error_sprintf() function.

errors.cpp: Renamed Fail() and Success() to isFailure()
and
isSuccess(), so that it's clearer to the reader that
they're boolean
tests.

errors.h: Match changes made to internal_error.cpp and
errors.cpp

internal_error.cpp: Renamed internal_error(const char
[], ...) to
internal_error_sprintf(). This fixes a link-time
conflict between
that function and internal_error(const char[]); the
signatures were
too close for the compiler to distinguish.

Also fixed bug in use of va_end() in
internal_error_sprintf(). This
would not have shown up except on certain odd
platforms; va_end() is
usually a no-op.

tcpUtilities.cpp: Added assertions and calls to
internal_error() to
confirm that Althea doesn't try to use SSL without
having SSL
compiled in. Like the standard assert() macro, these
are disabled
by compiling code with -DNDEBUG.

Modified ReadFromSocket() to clear up a
possibly-used-without-set warning.
Added comments.

SelectFolder.cpp: More use of enum errors instead of
int error codes.
Use assertions to catch any cases of what the
compiler warned of as
a possibly-used-before-set bug,

SelectFolder.h: Change function prototypes to reflect
changes to SelectFolder.cpp.

Server.cpp: Used #if 0 to comment out incomplete code
that is never
used. (The incomplete code justifiably triggers
compiler warnings.)

load_config.cpp: Used #if 0 to comment out unused
code.

load_config.h: #if 0 to comment out unused function
declaration.

Makefile: Added a useful development target to the
Makefile,
.cpp.i. This lets us generate .i files containing
preprocessed versions of the source.

support.cpp: New function, error_sprintf(). Formats an
error message
and then displays it by calling the existing error()
function.

support.h: Added prototype to match addition of
error_sprintf() to support.cpp.

More error handling

Steven Augart — Mon, 04 Feb 2002 23:17:54 -0000

Summary:
This patch adds error handling and debugging
routines that are useful
for general robustness.

To unpack:
patch --directory=${althea_source_root} --unified
--forward --strip=1 --backup --ignore-whitespace
more-error-handling.patch

Description:
This patch adds error handling and debugging routines
that are useful
for general robustness.

I have included files ending in .msg, for each changed
source file. These are intended for the CVS log. You
can delete them
afterwards. Thus, you can commit the changed version
of errors.h with:
cvs commit -F errors.h.msg errors.h && rm errors.h.msg

internal_error.cpp: Add routines to handle internal
errors.
The internal_error() functions are for Ťcan't
happenť conditions.

errors.cpp: Adds functions named Success() and Fail(),
for testing
Ťenum errorsť error codes. This abstracts various
ad_hoc tests
such as: (error_code < BEG_FAIL) or (error_code <
END_SUCCESS)
(Both of these forms are currently found in the code!)

errors.h: Added comments.
Added prototypes for new functions, Success(),
Fail(), and the
internal_error() family.

A temporary hack called XXX_CAST_INT_TO_ENUM_ERRORS
is also
here. The hack will go away once we've converted Althea
to handle all error codes as the Ťenum errorsť type
instead
of as the Ťintť type.

Makefile: Added rules for errors.o and
internal_error.o. Also
added a general CXXFLAGS definition, useful for implicit
.cpp ==> .o rules.

Makefile.debugging-warnings:
Added more comments.
Now turns on more warnings specific to the g++ compiler.
Now uses code generation options that may make it easier
to debug inline functions.

Message.cpp:
Began the conversion from integer error return codes to
Ťenum errorsť error return codes.

load_config.cpp:
Added some innocuous debugging code to help track
down a problem
that happens in the get_mail() function call while
load_Server() is running.

support.cpp:
In error(), all error messages now go to standard
error as well
as to a pop-up window. This is the case if
ALTHEA_DUMP_ERRORS_TO_STANDARD_ERR_TOO is defined.
This is part of an
attempt to track down a bug where the error windows
sometimes appear on
my display with no text (althea 0.5.7, gtk 1.2.10,
glibc 2.2.4, and
GNU g++ and libstdc++ version 2.95.3).

Makefile debugging warnings

Steven Augart — Sat, 02 Feb 2002 09:22:00 -0000

*Creates a new file, Makefile.debugging-warnings. By
uncommenting a
line in the master Makefile,
Makefile.debugging-warnings will be included
and a developer can use the warnings to check code.

*This patch was made against a version of Makefile
to which previously submitted patches were applied. It
is easy to adapt,
though, if those patches haven't been applied yet.