On Mi, 30.03.2005, 23:04 +0100 Paul Millar wrote in wine-patches:
I've tested this with an in-tree and out-of-tree compilation. Seems to work OK for me in both cases.
I'm working at wineprefixcreate also and while looking at your Patch, i found a race-condition when the user started wineprefixcreate:
wine.inf is searched first in "$datadir" and then in "$srcdatadir".
"$datadir" is "from the binary wine-tree" when wineprefixcreate was called with "--use-wine-tree" but it could also come from "the installed wine-version", when the option "--use-wine-tree" was not used.
wine.inf could be from a different wine-version in this case!
my modifications are:
before parsing the options:
+ topdir="" + srcdatadir=""
while parsing "--use-wine-tree": - datadir="$topdir/tools"
and later:
# Copy the .inf script and run it if [ -n "$topdir$srcdatadir" ] then for dir in "$topdir" "$srcdatadir"; do if [ -n "$dir" -a -e "$dir/tools/wine.inf" ]; then cp "$dir/tools/wine.inf" "$CROOT/windows/inf/wine.inf" break fi else cp "$datadir/wine.inf" "$CROOT/windows/inf/wine.inf" fi ----------
I moved srcdatadir one dir-level up to use the same level as "$topdir" and to make my life more easy, because i need "$srcdatadir/documentation/samples/config" for my patches.
Sorry, no diff here because I changed much more in my version. (Bug 2844 and Bug 2845).
-- by by ... ... Detlef
On Sun, Apr 10, 2005 at 11:46:38AM +0200, Detlef Riekenberg wrote:
Sorry, no diff here because I changed much more in my version.
Try to separate the changes, and submit a diff. I'm afraid things will go to /dev/null without one.
Hi Detlef,
Thanks for looking at this, but I'm afraid I'm slightly confused by your modified version of the patch ....
On Sunday 10 April 2005 10:46, Detlef Riekenberg wrote:
I'm working at wineprefixcreate also and while looking at your Patch, i found a race-condition when the user started wineprefixcreate:
wine.inf is searched first in "$datadir" and then in "$srcdatadir".
"$datadir" is "from the binary wine-tree" when wineprefixcreate was called with "--use-wine-tree" but it could also come from "the installed wine-version", when the option "--use-wine-tree" was not used.
wine.inf could be from a different wine-version in this case!
True, but this is the desirable effect, if I've understood correctly.
If the user specifies --use-wine-tree=/some/path/to/wine, then this is where she wants to get her wine.inf file from (assuming in-tree compilation).
The situation might arise if the user installs a release version of wine, but is trying out a CVS version to see if it fixes a problem. The user specifies a WINEPREFIX of ~/.wine-cvs (for example) whilst trying out the cvs version. Under these circumstances the user would want to generate the default registry using wine.inf from the CVS tree (i.e. from --use-wine-tree), and not the older values from the release wine.inf file in /usr/local (or wherever it was installed).
I can't see under what circumstance a user would specify --use-wine-tree and want to use the older, installed version of wine.inf. When would this be useful?
For out-of-tree compilation, the --use-source-tree option would be used in conjunction with the --use-wine-tree option, giving the correct result.
So, sorry I don't see the problem here. Could you explain further?
- topdir=""
- srcdatadir=""
AFAIK, this is basically a no-op unless you're up to mischief.
Under normal BASH shell expansion, there isn't a difference between an unset variable and one being empty:
paul@scaramanga:~$ unset a paul@scaramanga:~$ eval '[ ! -z "$a" ] && echo $a is $a' paul@scaramanga:~$ a="" eval '[ ! -z "$a" ] && echo $a is $a' paul@scaramanga:~$ a="some value" eval '[ ! -z "$a" ] && echo $a is $a' $a is some value
(the "eval"s are just to prevent the shell expanding the variable a before the initial assignment in the command line)
One *can* determin if a variable is set, but to me this would come under the category of "dubious activity"
paul@scaramanga:~$ eval 'set|grep -q "^a=" && echo $a is $a' paul@scaramanga:~$ a="" eval 'set|grep -q "^a=" && echo $a is $a' $a is paul@scaramanga:~$ a="some value" eval 'set|grep -q "^a=" && echo $a is $a' $a is some value
# Copy the .inf script and run it if [ -n "$topdir$srcdatadir" ] then for dir in "$topdir" "$srcdatadir"; do if [ -n "$dir" -a -e "$dir/tools/wine.inf" ]; then cp "$dir/tools/wine.inf" "$CROOT/windows/inf/wine.inf" break fi else cp "$datadir/wine.inf" "$CROOT/windows/inf/wine.inf" fi
Both $topdir and $srcdatadir are absolute directories, so the combination $topdir$srcdatadir doesn't make any sense. I *think* what you mean is something more like if [ -n "$topdir" -o -n "$srcdatadir" ]; then
I'd have slight issues with this generally as your duplicating code. OK, its just the "cp" here, but its bad form and makes code more error prone.
The duplication arises because of the previous issues with $datadir (which I think is a mistake).
I moved srcdatadir one dir-level up to use the same level as "$topdir" and to make my life more easy, because i need "$srcdatadir/documentation/samples/config" for my patches.
OK, but this does induce the code duplication above.
Perhaps, if you need to record the source-top-dir it would be better to define a $srctopdir inside the switch statement, defining $srcdatadir as $srctopdir/tools
I also noticed you made some cosmetic changes to winewrapper, but included this chunk:
@@ -106,7 +106,7 @@
if [ -z "$WINEPREFIX" -a ! -d "$HOME/.wine" ] then - if [ ! -z "$srcdir" ]; then + if [ -n "$srcdir" ]; then "$topdir/tools/wineprefixcreate" --update --use-wine-[...] else "$topdir/tools/wineprefixcreate" --update --use-wine-[...]
(kmail line wrapping be damned!)
Isn't ``[ ! -z "$var" ]'' equivalent to ``[ -n "$var" ]'' ?
Sorry, lots of questions ...
Paul.
Hi Paul.
Thanks for looking at this, but I'm afraid I'm slightly confused by your modified version of the patch ....
After thinking about my way the next day, i made a mistake:
My way was: - compiling "out-of-tree" - wineprefixcreate --use-wine-tree /path/to/wine.bin (but forgot --use-source-tree <dir>) - then "$datadir" give you the path to the local copy of "wine.inf"
But the last line was my mistake, because "$datadir" was changed inside "--use-wine-tree".
When the user forgot --use-source-tree, then "wine.inf" is not copied and your error-message comes up (.. bug in wine or it's packaging...)
The only additional validation, which might be useful, is to alert the user, when "--use-source-tree" is given, but "--use-wine-tree" is missing.
So, sorry I don't see the problem here. Could you explain further?
Afterwards, same for me... Sorry for that.
- topdir=""
- srcdatadir=""
AFAIK, this is basically a no-op unless you're up to mischief.
When the variable is set in the environment while starting wineprefixcreate, the script my use some files from somewhere in the universe.
# Copy the .inf script and run it if [ -n "$topdir$srcdatadir" ]
I *think* what you mean is something more like if [ -n "$topdir" -o -n "$srcdatadir" ]; then
yes, only in a short form.
The duplication arises because of the previous issues with $datadir (which I think is a mistake).
agree.
Perhaps, if you need to record the source-top-dir it would be better to define a $srctopdir inside the switch statement, defining $srcdatadir as $srctopdir/tools
of course, that's another way of "do the same"
I also noticed you made some cosmetic changes to winewrapper, but included this chunk:
- if [ ! -z "$srcdir" ]; then
- if [ -n "$srcdir" ]; then
Isn't ``[ ! -z "$var" ]'' equivalent to ``[ -n "$var" ]'' ?
I only want remove the negation, because there is a simple switch for the test, we need here.
-
Detlef Riekenberg -
Dimitrie O. Paun -
Paul Millar