binutils-gdb/gdb/doc/snapshots.readme

237 lines
10 KiB
Plaintext

GDB SNAPSHOT SYSTEM
(general info)
Updated 5/24/93
WHAT ARE GDB SNAPSHOTS
----------------------
Snapshots are an "image" of the main GDB development tree, captured at a
particular random instant in time. When you use the snapshots, you should
be able to maintain a local copy of GDB that is no more than one day older
than the official source tree used by the GDB maintainers.
The primary purpose of providing snapshots is to widen the group of
motivated developers that would like to help test, debug, and enhance GDB,
by providing you with access to the "latest and greatest" source.
This has several advantages, and several disadvantages.
First the advantages:
o Once we have a large base of motivated testers using the snapshots,
this should provide good coverage across all currently supported
GDB hosts and targets. If a new bug is introduced in GDB due to
fixing another bug or ongoing development, it should become
obvious much more quickly and get fixed before the next general
net release. This should help to reduce the chances of GDB being
released to the general public with a major bug that went unnoticed
during the release cycle testing because they are machine dependent.
We hope to greatly improve GDB's stability and reliability by
involving more people and more execution environments in the
prerelease testing.
o With access to the latest source, any diffs that you send to fix
bugs or add new features should be much easier for the GDB team
to merge into the official source base (after suitable review
of course). This encourages us to merge your changes quicker,
while they are still "fresh".
o Once your diffs are merged, you can obtain a new copy of GDB
containing your changes almost immediately. Thus you do not
have to maintain local copies of your changes for any longer
than it takes to get them merged into the official source base.
This encourages you to send in changes quicker.
And the disadvantages:
o The snapshot you get will be largely untested and of unknown quality.
It may fail to configure or compile. It may have serious bugs.
You should always keep a copy of the last known working version
before updating to the current snapshot, or at least be able to
regenerate a working version if the latest snapshot is unusable
in your environment for some reason.
If a production version of GDB has a bug and a snapshot has the fix,
and you care about stability, you should put only the fix for that
particular problem into your production version. Of course, if you
are eager to test GDB, you can use the snapshot versions in your
daily work, but users who have not been consulted about whether they
feel like testing GDB should generally have something which is at
least as bug free as the last released version.
o Providing timely response to your questions, bug reports, and
submitted patches will require the GDB development team to allocate
time from an already thin time budget. Please try to help us make
this time as productive as possible. See the section below about
how to submit changes.
HOW TO GET THE SNAPSHOTS
------------------------
The current plan is to provide a full snapshot once weekly, and incremental
diffs on a daily basis. Each daily diff will be relative to the source
tree for the previous day after applying all incremental diffs to date.
The files will be available via anonymous ftp from ftp.cygnus.com, in
directory pub/gdb, and should look something like:
gdb-930401.tar.z
gdb-930401-930402.diff.z
gdb-930402-930403.diff.z
gdb-930403-930404.diff.z
.
.
.
At some point, the files should automatically appear during the evening
as a result of an automatically run process each evening. For the moment
however, the process will be manually run by one of the gdb maintainers
and the appropriate files moved to the ftp area at some convenient point
during the day.
Note that the current plan is to provide GNU gzip compressed files only.
You can ftp gzip from prep.ai.mit.edu in directory pub/gnu.
Also, as the gcc developers did with their gcc snapshot system, even though
we will make the snapshots available on a publically accessible ftp area,
we ask that recipients not widely publicise their availability. The motivation
for this request is not to hoard them, but to avoid the situation where
the general GDB user base naively attempts to use the snapshots, has trouble
with them, complains publically, and the reputation of GDB declines because
of a perception of instability or lack of quality control.
GDB TEST SUITE
--------------
A test suite is distributed as an integral part of the snapshots. However,
to use it you will need to get a copy of the dejagnu testing framework.
Snapshots of dejagnu are available alongside the GDB snapshots, using
the same naming conventions as the GDB snapshots. Once you have installed
the dejagnu framework, a simple "make check" in the GDB directory should
be sufficient to run the tests.
Note that the test suite is still in its infancy. The test framework
itself might not install on your system if you have an environment that
is not similar to one that the GDB developers already use. The tests
themselves only cover a small portion of GDB features, and what tests
do exist for a feature are not exhaustive. New tests are welcomed.
GETTING HELP, GDB DISCUSSIONS, etc
----------------------------------
Mail sent to gdb-testers@cygnus.com goes to everyone on the list of gdb
testers, which should include everyone getting the gdb snapshots. It is
appropriate whenever you wish your mail to be seen by all the testers.
This would include announcements of any kind, notices of intent to implement
a specific enhancement (to coordinate with other people on the list), etc.
Before sending something to gdb-testers, ask yourself if what you are about
to send would be something you would care to see show up in your mailbox if
it was sent by someone else.
Mail sent to gdb-patches@cygnus.com goes to gdb support people internal to
Cygnus. Despite the name, it is appropriate for more than just patches.
Questions about the snapshots, problems accessing the snapshots, bug reports
without patches, requests for advice on how to track down a bug you have
encountered, discussion about bug fixes or enhancements in progress, etc are
all welcome in gdb-patches. Usually mail sent to gdb-patches will result in
a short private email discussion between you and one or more of the gdb
developers who can assist you with simple questions or handle your patches.
Note that gdb-patches is *not* a general gdb electronic support line.
If you are in need of such support, you probably should not be using the
snapshots and should seek out one of the commercial suppliers of support
for free software.
Do *not* send any questions about the snapshots or patches specific to
the snapshots to bug-gdb@prep.ai.mit.edu (gateway'd to the usenet group
gnu.gdb.bug). Nobody there will have any idea what you are talking about
and it will just cause confusion.
BUG REPORTS
-----------
Send bug reports to gdb-patches@cygnus.com.
Note that since no testing is done on the snapshots, and snapshots may even
be made when gdb is in an inconsistent state, it may not be unusual for an
occasional snapshot to have a very obvious bug, such as failure to compile
on *any* machine. It is likely that such bugs will be fixed by the next
snapshot, so it really isn't necessary to report them unless they persist
for a couple days.
Bugs which are non-obvious, such as failure to compile on only a specific
machine, a new machine dependent bug (particularly one not detected by the
testsuite), missing machine dependent files, etc should be reported when
you discover them, or have a suggested patch to fix them.
FORMAT FOR PATCHES
------------------
If you have a fix for a bug, or an enhancement to submit, send your
patch to gdb-patches@cygnus.com. Here are some simple guidelines for
submitting patches:
o Use "context diffs" for patches. A typical command for generating
context diffs is "diff -rc gdb-old gdb-new".
o Use the "minimalist approach" for patches. That is, each patch
should address only one particular bug, new feature, etc. Do not
save up many unrelated changes and submit them all in one big
patch, since in general, the larger the patch the more difficult
it is for us to decide if the patch is either correct or
desirable. And if we find something about the patch that needs
to be corrected before it can be installed, we would have to reject
the entire patch, which might contain changes which otherwise would
be accepted if submitted separately.
o Submit a sample ChangeLog entry with your patch. See the existing
GDB ChangeLog for examples of what a ChangeLog entry should look
like. The emacs command ^X4A will create a ChangeLog entry header
for you.
BISON and BYACC
---------------
GDB's language parsers are all portable, and can be compiled with bison,
byacc, traditional Unix yacc, or other compatible parser generators.
For various reasons, Cygnus uses byacc rather than bison by default. When
a general gdb distribution is made, this default is switched back to bison.
The snapshots follow the Cygnus default. Your options, if you do not already
have byacc installed, include:
o Hack the upper level Makefile.in lines that look like:
BISON = `if [ -f $${rootme}/byacc/byacc ] ; \
then echo $${rootme}/byacc/byacc ; \
else echo byacc ; \ <== change
fi`
to replace "byacc" with either "yacc" or "bison -y".
o Fetch the byacc snapshot from the same location as the gdb snapshots
and install byacc.
o Specify BISON=yacc on the make command line to override the default.
UNIX MAKE and GNU MAKE
----------------------
When you build gdb in the same directory as the source, you should be able
to use any available "make" that has traditional UNIX make functionality.
If you build gdb in a separate directory tree from the source, using the
configure "--srcdir" option, then only GNU make is fully supported, although
other makes with complete VPATH support should work (SunOS make for example).
Thanks for your help and support.
-Fred Fish
Cygnus Support