Re: [PATCH]: mknetrel builds Guile #2: debug
Christopher Faylor [EMAIL PROTECTED] writes: I've applied a slight variation of this patch. I kept the exit in usage but made it exit 1 when usage was given an argument. Thanks. Maybe such debugging is overkill, but this kind of thing helped me a lot during the porting/patching of Guile and libtool. Jan. -- Jan Nieuwenhuizen [EMAIL PROTECTED] | GNU LilyPond - The music typesetter http://www.xs4all.nl/~jantien | http://www.lilypond.org
Re: [PATCH]: mknetrel builds Guile #2: debug
Christopher Faylor [EMAIL PROTECTED] writes:
- add debugging feature: drop to shell upon failure
I normally run mknetrel in the background. In fact, I just added a
couple of options to accommodate this. They would conflict with this
patch.
Ok.
If you want to add this functionality, introduce an option. I don't
want to have this as the default.
Sure, that's inconvenient. I thought about adding it to -x, but you
may well want to have a detailed, noninteractive log, at times. So,
I decided to add -X option for this.
Also, I made -h exit successfully, not to annoy you or anything, saw
you were messing with this too. Hope you like the way I did it. I
thought about adding a help () funtion, but decided that would be
overkill.
What I really think we should also do, is to start the script with
set -e
and let bash decide when something's wrong. Then we can rid of stuff
like:
mkdir foo || exit 1
It is annoying that you'll have to add '|| true' clauses to some
checks that are allowed to fail, but in general, I feel that a bourn
script *should* start with set -e. Any sane interpreter stops upon
error. What do you think?
Greetings,
Jan.
Btw: A very silly question, but I can't seem to get `cvs diff' to
respect your (Cygwin's?) wishes of excluding the ChangeLog from
the diff. It doesn't grok -X,--exclude options, most annoying.
How do you manage?
2002-07-09 Jan Nieuwenhuizen [EMAIL PROTECTED]
* bin/mknetrel: Invoke atexit () before exiting.
(atexit): New function.
(drop): New funtion.
(usage): Return (previously: exit unsuccessfully).
(setup): Add -X option to make mknetrel drop to shell upon error.
Exit successfully for -h, unsuccessfully upon usage error.
(setup1): Exit unsuccessfully upon usage error.
* .cvsignore: Add .bash_mknetrel.
? pats
Index: .cvsignore
===
RCS file: /cvs/cygwin-apps/mknetrel/.cvsignore,v
retrieving revision 1.2
diff -p -u -r1.2 .cvsignore
--- .cvsignore 9 Jul 2002 03:32:14 - 1.2
+++ .cvsignore 9 Jul 2002 12:06:38 -
@@ -4,6 +4,7 @@ Makefile
#*
*.swp
.cvsignore
+.bash_mknetrel
build
inst
log
Index: ChangeLog
===
RCS file: /cvs/cygwin-apps/mknetrel/ChangeLog,v
retrieving revision 1.14
diff -p -u -r1.14 ChangeLog
--- ChangeLog 9 Jul 2002 06:01:59 - 1.14
+++ ChangeLog 9 Jul 2002 12:06:38 -
@@ -1,3 +1,15 @@
+2002-07-09 Jan Nieuwenhuizen [EMAIL PROTECTED]
+
+ * bin/mknetrel: Invoke atexit () before exiting.
+ (atexit): New function.
+ (drop): New funtion.
+ (usage): Return (previously: exit unsuccessfully).
+ (setup): Add -X option to make mknetrel drop to shell upon error.
+ Exit successfully for -h, unsuccessfully upon usage error.
+ (setup1): Exit unsuccessfully upon usage error.
+
+ * .cvsignore: Add .bash_mknetrel.
+
2002-07-09 Christopher Faylor [EMAIL PROTECTED]
* bin/mknetrel (doconfig): Fix more quoting problems.
Index: bin/mknetrel
===
RCS file: /cvs/cygwin-apps/mknetrel/bin/mknetrel,v
retrieving revision 1.33
diff -p -u -r1.33 mknetrel
--- bin/mknetrel9 Jul 2002 06:32:17 - 1.33
+++ bin/mknetrel9 Jul 2002 12:06:38 -
@@ -32,6 +32,24 @@ read_user_config() {
export knownpath
}
+atexit () { :; }
+
+trap atexit 0 9 15
+
+# Interactive shell with state of bash at exit upon failure
+drop ()
+{
+[ $? -eq 0 ] trap - 0 9 15 exit 0 || true
+echo $0: dropping to shell...
+set | sed -e 's/^\(BASH_VERSINFO=\)/#\1/' \
+ -e 's/^\(EUID=\)/#\1/' \
+ -e 's/^\(PPID=\)/#\1/' \
+ -e 's/^\(SHELLOPTS=\)/#\1/' \
+ -e 's/^\(UID=\)/#\1/' .bash_mknetrel
+echo PS1='mknetrel-debug \w$ ' .bash_mknetrel
+exec bash --rcfile .bash_mknetrel
+}
+
setvars() {
cd $netrel_root || exit 1
here=`pwd`
@@ -86,19 +104,20 @@ setup() {
read_user_config
clean=false
opt=-O2
-while getopts 'ihtbcCgNnxSoB' o; do
+while getopts 'ihtbcCgNnxXSoB' o; do
case $o in
'b') rebuild() { dorebuild; } ;;
'B') background() { return 0;}; output() { return 0;} ;;
'c') config() { doconfig; }; rebuild() { dorebuild; } ;;
'C') clean=: ;;
'g') opt=-O2 -g; rebuild() { dorebuild; } ;;
- 'h') usage ;;
+ 'h') usage; exit 0;;
'i') knownpath= ;;
'n') mkdist() { :; } ;;
'o') output() { return 0; } ;;
'S') stripbins() { return 1; } ;;
'x') verbose() { set -x; } ;;
+ 'X') atexit() { drop; } ;;
'?') exit 1 ;;
esac
done
@@ -107,6 +126,7 @@ setup() {
what=$1
if [ -z $what ]; then
usage missing package name
+ exit 1
Re: [PATCH]: mknetrel builds Guile #2: debug
Jan Nieuwenhuizen wrote: Btw: A very silly question, but I can't seem to get `cvs diff' to respect your (Cygwin's?) wishes of excluding the ChangeLog from the diff. It doesn't grok -X,--exclude options, most annoying. How do you manage? Remove the ChangeLog diff from the patch. Copy the text from the ChangeLog to the top of the diff file. Patch is smart enough to ignore the non-diff text in the front of the file. Earnie.
Re: [PATCH]: mknetrel builds Guile #2: debug
Earnie Boyd [EMAIL PROTECTED] writes: Btw: A very silly question, but I can't seem to get `cvs diff' to respect your (Cygwin's?) wishes of excluding the ChangeLog from the diff. It doesn't grok -X,--exclude options, most annoying. How do you manage? Remove the ChangeLog diff from the patch. Copy the text from the ChangeLog to the top of the diff file. Patch is smart enough to ignore the non-diff text in the front of the file. Hrmm. If you're going to have nonstandard requests, it would be handy if these somehow would be handled automagically. How dead is CVS development, maybe they would take a patch? Greetings, Jan -- Jan Nieuwenhuizen [EMAIL PROTECTED] | GNU LilyPond - The music typesetter http://www.xs4all.nl/~jantien | http://www.lilypond.org
Re: [PATCH]: mknetrel builds Guile #2: debug
Jan Nieuwenhuizen wrote: Earnie Boyd [EMAIL PROTECTED] writes: Btw: A very silly question, but I can't seem to get `cvs diff' to respect your (Cygwin's?) wishes of excluding the ChangeLog from the diff. It doesn't grok -X,--exclude options, most annoying. How do you manage? Remove the ChangeLog diff from the patch. Copy the text from the ChangeLog to the top of the diff file. Patch is smart enough to ignore the non-diff text in the front of the file. Hrmm. If you're going to have nonstandard requests, it would be handy if these somehow would be handled automagically. How dead is CVS development, maybe they would take a patch? This has nothing to do with CVS. It has to do with standards set by the development teams and the fact that your changes to ChangeLog often (most likely) cause merge conflicts. Earnie.
Re: [PATCH]: mknetrel builds Guile #2: debug
Charles Wilson [EMAIL PROTECTED] writes: That's why I am investigating transitioning us to a cygwin port of cvsnt (no, it's not just for native windows anymore...) And it does work with emacs' pcvs.el, right? The thing about the ChangeLog stuff, is that it really isn't that non-standard. MOST projects dislike *patches* to ChangeLog files, Maybe just my ignorance that I haven't seen these requests before. It's not so much that I dislike project specific stuff (look at LilyPond), but the fact that the current tools don't handle it gracefully, let alone automagically. This is because a patch to ChangeLog *never* (okay, ALMOST never) applies cleanly, Ack. I tend NOT to modify the ChangeLog at all -- I put my ChangeLog entries in a separate file, and paste that into the email by hand... Ok, I'm going to try that. Jan. -- Jan Nieuwenhuizen [EMAIL PROTECTED] | GNU LilyPond - The music typesetter http://www.xs4all.nl/~jantien | http://www.lilypond.org
Re: [PATCH]: mknetrel builds Guile #2: debug
Jan Nieuwenhuizen wrote: That's why I am investigating transitioning us to a cygwin port of cvsnt (no, it's not just for native windows anymore...) And it does work with emacs' pcvs.el, right? Dunno. It's based on the cvs code; most of the changes are (a) bugfixes -- what a concept (b) text/binary handling (c) the daemon (yay!) Anyway, nothing's going to happen until at least September, and even then, there will be a long and thorough testing period before we change *anything*. cvs is core; I understand the need for caution. --Chuck
Re: [PATCH]: mknetrel builds Guile #2: debug
mmm, time for arch/subversion/another one of the recent pack of opensource / free software revision control systems? Rob - Original Message - From: Charles Wilson [EMAIL PROTECTED] To: Jan Nieuwenhuizen [EMAIL PROTECTED] Cc: Earnie Boyd [EMAIL PROTECTED] Sent: Wednesday, July 10, 2002 12:42 AM Subject: Re: [PATCH]: mknetrel builds Guile #2: debug Jan Nieuwenhuizen wrote: Hrmm. If you're going to have nonstandard requests, it would be handy if these somehow would be handled automagically. How dead is CVS development, maybe they would take a patch? Deader than roadkill, IMO. I think they are in permanennt bugfix only mode (except for new features developed in house by the maintainers themselves). That's why I am investigating transitioning us to a cygwin port of cvsnt (no, it's not just for native windows anymore...) The thing about the ChangeLog stuff, is that it really isn't that non-standard. MOST projects dislike *patches* to ChangeLog files, and request that ChangeLog entries be sent in cut-n-paste form. This is because a patch to ChangeLog *never* (okay, ALMOST never) applies cleanly, since it is extremely likely that your patch will be queued up behind others, which ALSO affect the ChangeLog at exactly the same spot yours does: the top. I tend NOT to modify the ChangeLog at all -- I put my ChangeLog entries in a separate file, and paste that into the email by hand... --Chuck
Re: [PATCH]: mknetrel builds Guile #2: debug
On Tue, Jul 09, 2002 at 02:45:46PM +0200, Jan Nieuwenhuizen wrote: 2002-07-09 Jan Nieuwenhuizen [EMAIL PROTECTED] * bin/mknetrel: Invoke atexit () before exiting. (atexit): New function. (drop): New funtion. (usage): Return (previously: exit unsuccessfully). (setup): Add -X option to make mknetrel drop to shell upon error. Exit successfully for -h, unsuccessfully upon usage error. (setup1): Exit unsuccessfully upon usage error. * .cvsignore: Add .bash_mknetrel. I've applied a slight variation of this patch. I kept the exit in usage but made it exit 1 when usage was given an argument. Thanks. cgf
[PATCH]: mknetrel builds Guile #2: debug
- add debugging feature: drop to shell upon failure
diff -purNX .cvsignore ../mknetrel.jcn1.init/.cvsignore ./.cvsignore
--- ../mknetrel.jcn1.init/.cvsignoreMon Jul 8 15:42:06 2002
+++ ./.cvsignoreMon Jul 8 16:04:09 2002
@@ -2,3 +2,4 @@ CVS
*~
#*
.#*
+.bash_mknetrel
diff -purNX .cvsignore ../mknetrel.jcn1.init/ChangeLog ./ChangeLog
--- ../mknetrel.jcn1.init/ChangeLog Tue Jul 9 01:17:39 2002
+++ ./ChangeLog Tue Jul 9 01:17:29 2002
@@ -1,5 +1,16 @@
2002-07-08 Jan Nieuwenhuizen [EMAIL PROTECTED]
+ * VERSION mknetrel.jcn2.debug
+
+ * .cvsignore: Add .bash_mknetrel.
+
+ * bin/mknetrel (drop): New funtion. Add trap command, to invoke
+ drop () upon unsuccessful exit, so that user make investigate at
+ the spot what went wrong. To test this, make a function such as
+ postconfig () or postbuild () return an error.
+ (usage): Don't drop to shell. Use exit value 2, commonly for
+ usage errors.
+
* VERSION mknetrel.jcn1.init
* .cvsignore: New file.
diff -purNX .cvsignore ../mknetrel.jcn1.init/bin/mknetrel ./bin/mknetrel
--- ../mknetrel.jcn1.init/bin/mknetrel Tue Jul 9 01:06:32 2002
+++ ./bin/mknetrel Tue Jul 9 01:07:05 2002
@@ -14,6 +14,23 @@ read_user_config() {
: ${build_config_opts='--build=i686-pc-linux --host=i686-pc-cygwin
--target=i686-pc-cygwin'}
}
+# Interactive shell with state of bash at exit upon failure
+drop ()
+{
+[ $? -eq 0 ] trap - 0 9 15 exit 0 || true
+echo $0: dropping to shell...
+set | sed -e 's/^\(BASH_VERSINFO=\)/#\1/' \
+ -e 's/^\(EUID=\)/#\1/' \
+ -e 's/^\(PPID=\)/#\1/' \
+ -e 's/^\(SHELLOPTS=\)/#\1/' \
+ -e 's/^\(UID=\)/#\1/' .bash_mknetrel
+echo PS1='mknetrel-debug \w$ ' .bash_mknetrel
+exec bash --rcfile .bash_mknetrel
+}
+
+# Call drop at exit
+trap drop 0 9 15
+
setvars() {
mknetrel=$(cd $(dirname $0)/..; pwd)
cd $netrel_root || exit 1
@@ -329,7 +346,7 @@ Options:
Available packages:
$(cd $(dirname $0)/../extra echo *[a-z])
12
-exit 1
+trap - 0 9 15 exit 2
}
rest() {
--
Jan Nieuwenhuizen [EMAIL PROTECTED] | GNU LilyPond - The music typesetter
http://www.xs4all.nl/~jantien | http://www.lilypond.org
