Re: [Monetdb-developers] [Monetdb-checkins] MonetDB/conf conf.bash, 1.129, 1.130
On Wed, Jul 05, 2006 at 07:34:03AM +0000, Stefan Manegold wrote:
[...]
Debian's default flex 2.5.31 does not like the Pathfinder code; hence, we use our own flex 2.5.33 instead.
[...]
I haven't realized that we have a problem there, but your commit message seems suspicious to me: Your checkin only modifies conf.bash. That means, only development folks will be affected by your checkin. Users will not have your specialized flex version installed. If we indeed have problems with flex 2.5.31, I guess we'll either have to either (a) change our code such that it does compile with 2.5.31. I don't know what the problem was on Debian, so I have no clue if we could fix this. Or (b) we should check the flex version in ./configure and abort compilation in case the installed flex version is not sufficient to compile Pathfinder. Maybe you can enlighten me (maybe also others) what the actual flex problem was. I cannot find anything in the TestWeb. Jens -- Jens Teubner Technische Universitaet Muenchen, Department of Informatics D-85748 Garching, Germany Tel: +49 89 289-17259 Fax: +49 89 289-17263 This email was written with 100% recycled electrons.
On 05-07-2006 10:13:15 +0200, Jens Teubner wrote:
On Wed, Jul 05, 2006 at 07:34:03AM +0000, Stefan Manegold wrote:
[...]
Debian's default flex 2.5.31 does not like the Pathfinder code; hence, we use our own flex 2.5.33 instead.
[...]
I haven't realized that we have a problem there, but your commit message seems suspicious to me:
Your checkin only modifies conf.bash. That means, only development folks will be affected by your checkin. Users will not have your specialized flex version installed.
If we indeed have problems with flex 2.5.31, I guess we'll either have to either [snip]
+1 It is the same as the bison situation on OSX, IMO. That said, IIRC this change shouldn't affect tarball users, as they don't need any of the generator tools, right?
On Wed, Jul 05, 2006 at 10:20:46AM +0200, Fabian Groffen wrote:
It is the same as the bison situation on OSX, IMO. That said, IIRC this change shouldn't affect tarball users, as they don't need any of the generator tools, right?
I might be wrong here, but I think OSX was a case of option (b). That is, we check for the `locations' feature in ./configure. Jens -- Jens Teubner Technische Universitaet Muenchen, Department of Informatics D-85748 Garching, Germany Tel: +49 89 289-17259 Fax: +49 89 289-17263 Computer Scientist, n.: A device that turns coffee into programs.
On Wed, Jul 05, 2006 at 10:13:15AM +0200, Jens Teubner wrote:
On Wed, Jul 05, 2006 at 07:34:03AM +0000, Stefan Manegold wrote:
[...]
Debian's default flex 2.5.31 does not like the Pathfinder code; hence, we use our own flex 2.5.33 instead.
[...]
I haven't realized that we have a problem there, but your commit message seems suspicious to me:
Your checkin only modifies conf.bash. That means, only development folks will be affected by your checkin. Users will not have your specialized flex version installed.
If we indeed have problems with flex 2.5.31, I guess we'll either have to either
(a) change our code such that it does compile with 2.5.31. I don't know what the problem was on Debian, so I have no clue if we could fix this.
Or
(b) we should check the flex version in ./configure and abort compilation in case the installed flex version is not sufficient to compile Pathfinder.
Maybe you can enlighten me (maybe also others) what the actual flex problem was. I cannot find anything in the TestWeb.
Check the "Debian3.1" column at http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/... or go directly to http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/... With flex 2.5.31 (at least on Debian; the only platform in the TestWebPool that has flex 2.5.31; all others have either flex 2.5.4 or 2.5.33), we get tons of ======== syntax error, unexpected at_StringLiteral on line 1 (next token is 'should work'') parse error: XQuery parsing failed ======== with pathfinder; with flex 2.5.33 these are gone (on Debian). Admittedly, my "fix" only works for the testweb, and indeed, if this is a flex 2.5.31 problem on other platforms as well, we should rather have configure check for the flex version, and quit in case we find 2.5.31. However, since I don't know, whether this in only a problem with the default flex 2.5.31 on our 64-bit Debian 3.1, I went for this quick hack. I also did not spend any time on analysing whether we could change our code to work properly with the default flex 2.5.31 on our 64-bit Debian 3.1. Indeed, only "development folks" that use conf/conf.bash on a Debian system will benefit from my "fix". However, the tarball come with the flex/bison generated C code, hence these tools should not be required for users; and hence, the problem should not occur for users. Stefan
Jens
-- Jens Teubner Technische Universitaet Muenchen, Department of Informatics D-85748 Garching, Germany Tel: +49 89 289-17259 Fax: +49 89 289-17263
This email was written with 100% recycled electrons.
Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Monetdb-developers mailing list Monetdb-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-developers
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
I just checked this with flex 2.5.31 on my 64-bit Fedora Core 4 desktop, and I get the same problems as on Debian; I.e., this indeed seems to be a general problem with flex 2.5.31 and our pathfinder lexer code. hence, unless we can fix our pathfinder lexer code to work properly with flex 2.5.31, we should indeed make configure check for the flex version, and quit once we find flex 2.5.31. ... I won't have time for this before next week ... Stefan
Check the "Debian3.1" column at http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/... or go directly to http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/...
With flex 2.5.31 (at least on Debian; the only platform in the TestWebPool that has flex 2.5.31; all others have either flex 2.5.4 or 2.5.33), we get tons of ======== syntax error, unexpected at_StringLiteral on line 1 (next token is 'should work'') parse error: XQuery parsing failed ========
with pathfinder; with flex 2.5.33 these are gone (on Debian).
Admittedly, my "fix" only works for the testweb, and indeed, if this is a flex 2.5.31 problem on other platforms as well, we should rather have configure check for the flex version, and quit in case we find 2.5.31. However, since I don't know, whether this in only a problem with the default flex 2.5.31 on our 64-bit Debian 3.1, I went for this quick hack.
I also did not spend any time on analysing whether we could change our code to work properly with the default flex 2.5.31 on our 64-bit Debian 3.1.
Indeed, only "development folks" that use conf/conf.bash on a Debian system will benefit from my "fix". However, the tarball come with the flex/bison generated C code, hence these tools should not be required for users; and hence, the problem should not occur for users.
Stefan
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
On 05-07-2006 11:29:44 +0200, Stefan Manegold wrote:
I.e., this indeed seems to be a general problem with flex 2.5.31 and our pathfinder lexer code.
hence, unless we can fix our pathfinder lexer code to work properly with flex 2.5.31, we should indeed make configure check for the flex version, and quit once we find flex 2.5.31.
I think "Debian stable" is a just out of our capacities to maintain. Since releases should work fine, add the configure check/requirement, like the bison check; only failing if not in a dist tarball.
On Wed, Jul 05, 2006 at 11:29:44AM +0200, Stefan Manegold wrote:
I just checked this with flex 2.5.31 on my 64-bit Fedora Core 4 desktop, and I get the same problems as on Debian;
I.e., this indeed seems to be a general problem with flex 2.5.31 and our pathfinder lexer code.
hence, unless we can fix our pathfinder lexer code to work properly with flex 2.5.31, we should indeed make configure check for the flex version, and quit once we find flex 2.5.31.
Hi Stefan, thanks for clarifying this. Reading your original checkin message, I thought that this was a *compilation* problem, which apparently it is not. The error message is strange. We haven't experienced any problems on that end so far, and I suspect this is really a flex problem/bug. So a ./configure check I guess is the right "solution". Jens P.S.: Stefan, much like you, I won't find time to look into this for at least a week. Whoever looks into this first just lets the other know, ok? -- Jens Teubner Technische Universitaet Muenchen, Department of Informatics D-85748 Garching, Germany Tel: +49 89 289-17259 Fax: +49 89 289-17263 In theory, practice and theory are the same, but in practice they are different. -- Larry McVoy
"Unexpectedly", I found/made some time, and fixed the flex issue "properly": - The tar-ball releases come with the bison/flex generated tools, hence, configure only needs to check for the respetive tool when compiling the CVS sources. - Pathfinder's configure (when compiling the CVS source) check's the flex version, and quit's in case it's 2.5.31, suggesting to install/use 2.5.4 or 2.5.33 instead. Stefan On Wed, Jul 05, 2006 at 11:45:44AM +0200, Jens Teubner wrote:
On Wed, Jul 05, 2006 at 11:29:44AM +0200, Stefan Manegold wrote:
I just checked this with flex 2.5.31 on my 64-bit Fedora Core 4 desktop, and I get the same problems as on Debian;
I.e., this indeed seems to be a general problem with flex 2.5.31 and our pathfinder lexer code.
hence, unless we can fix our pathfinder lexer code to work properly with flex 2.5.31, we should indeed make configure check for the flex version, and quit once we find flex 2.5.31.
Hi Stefan,
thanks for clarifying this. Reading your original checkin message, I thought that this was a *compilation* problem, which apparently it is not.
The error message is strange. We haven't experienced any problems on that end so far, and I suspect this is really a flex problem/bug. So a ./configure check I guess is the right "solution".
Jens
P.S.: Stefan, much like you, I won't find time to look into this for at least a week. Whoever looks into this first just lets the other know, ok?
-- Jens Teubner Technische Universitaet Muenchen, Department of Informatics D-85748 Garching, Germany Tel: +49 89 289-17259 Fax: +49 89 289-17263
In theory, practice and theory are the same, but in practice they are different. -- Larry McVoy
Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Monetdb-developers mailing list Monetdb-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-developers
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
participants (3)
-
Fabian Groffen -
Jens Teubner -
Stefan Manegold