Re: new expr(1) behaviour breaks libtool
I have just committed code to expr which will cause it to behave more like the old expr did in the presence of an EXPR_COMPAT environment variable. Ports can then be set up to build with this variable defined until the libtool maintainers fix up their act. Thanks, with this and some patches to the ghostscript-gnu and jade ports, I can now build releases with docs. John -- John Hay -- [EMAIL PROTECTED] / [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: new expr(1) behaviour breaks libtool
John Hay wrote: I see the new new behaviour of expr(1) requires you to add '--' if your commandline arguments might start with a '-'. This does break things a little because our old expr(1) does not understand a '--' in the beginning and the new one don't work right without it. :-((( I'm almost positive this issue was discussed before. Check the follow ups to the commit. The only one I could find was in -current, where Kris asked if w3m or expr is to blame and Garrett said w3m is to blame. Garrett is right. The expr is being called with non-POSIX arguments by the w3c scripts. H. I can understand the requirement to eat '--', but to throw a tantrum just because the commandline started with a '-' is a little too much. BTW, was the response posted somewhere? I searched through -standards, -commit and -current but couldn't find it. Maybe I just didn't ask the right question to the search engine or maybe it was in another list. The correct way to handle the continuing brokenness is to provide patches that will get applied to libtool. Also, you complaint about different versionsion of FreeBSD being different is wrong: POSIX compliant scripts will call the command with POSIX compliant arguments, regardless of the FreeBSD version. Just because you can get away with it on older versions of FreeBSD doesn't mean that the patches shouldn't be applied there, too: they should, because it's important that the resulting scripts be standards compliant, even if the particular version of FreeBSD you are using has an implementation of expr that illegally extends the standard. -- Terry To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: new expr(1) behaviour breaks libtool
On Sun, 21 Apr 2002 20:17:29 +0200 (SAT), John Hay [EMAIL PROTECTED] said: expr -lgrove : -l\(.*\) expr -- -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) If we are going to leave this behaviour, we will have to teach libtool how to call expr(1) differently on -stable and -current and it looks like yet again different from the rest of the world. :-((( No, you just have to teach libtool to properly protect the arguments to expr(1). libtool 1.4 does this for some numeric operations, but not for string parsing. expr(1) goes into some detail about this, but the simple answer is: expr X$arg : 'X-l\(.*\)' is what's required. (There is nothing special about the character `X', other than it not being a hyphen.) I have just committed code to expr which will cause it to behave more like the old expr did in the presence of an EXPR_COMPAT environment variable. Ports can then be set up to build with this variable defined until the libtool maintainers fix up their act. -GAWollman To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
new expr(1) behaviour breaks libtool
Hi, I see the new new behaviour of expr(1) requires you to add '--' if your commandline arguments might start with a '-'. This does break things a little because our old expr(1) does not understand a '--' in the beginning and the new one don't work right without it. :-((( The place where I noticed it was when libtool started to complain when compiling jade. Libtool does things like: expr -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) expr -lsp : -l\(.*\) expr -lm : -l\(.*\) expr -lgrove : -l\(.*\) On -current this now have to be: expr -- -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) expr -- -lsp : -l\(.*\) expr -- -lm : -l\(.*\) expr -- -lgrove : -l\(.*\) If we are going to leave this behaviour, we will have to teach libtool how to call expr(1) differently on -stable and -current and it looks like yet again different from the rest of the world. :-((( Yes, I did read the commit message, but I still think the behaviour of the new expr(1) is wrong. John -- John Hay -- [EMAIL PROTECTED] / [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: new expr(1) behaviour breaks libtool
John Hay [EMAIL PROTECTED] writes: I see the new new behaviour of expr(1) requires you to add '--' if your commandline arguments might start with a '-'. This does break things a little because our old expr(1) does not understand a '--' in the beginning and the new one don't work right without it. :-((( I'm almost positive this issue was discussed before. Check the follow ups to the commit. The place where I noticed it was when libtool started to complain when compiling jade. Libtool does things like: expr -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) expr -lsp : -l\(.*\) expr -lm : -l\(.*\) expr -lgrove : -l\(.*\) On -current this now have to be: expr -- -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) expr -- -lsp : -l\(.*\) expr -- -lm : -l\(.*\) expr -- -lgrove : -l\(.*\) If we are going to leave this behaviour, we will have to teach libtool how to call expr(1) differently on -stable and -current and it looks like yet again different from the rest of the world. :-((( This should exactly match the behavior of any certified UNIX system. Yes, I did read the commit message, but I still think the behaviour of the new expr(1) is wrong. Not according to the Standard, or the response from Garrett's request for clarification of the Standard. Best regards, Mike Barcroft To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: new expr(1) behaviour breaks libtool
I see the new new behaviour of expr(1) requires you to add '--' if your commandline arguments might start with a '-'. This does break things a little because our old expr(1) does not understand a '--' in the beginning and the new one don't work right without it. :-((( I'm almost positive this issue was discussed before. Check the follow ups to the commit. The only one I could find was in -current, where Kris asked if w3m or expr is to blame and Garrett said w3m is to blame. The place where I noticed it was when libtool started to complain when compiling jade. Libtool does things like: expr -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) expr -lsp : -l\(.*\) expr -lm : -l\(.*\) expr -lgrove : -l\(.*\) On -current this now have to be: expr -- -L/export/ports/textproc/jade/work/jade-1.2.1/lib/.libs : -l\(.*\) expr -- -lsp : -l\(.*\) expr -- -lm : -l\(.*\) expr -- -lgrove : -l\(.*\) If we are going to leave this behaviour, we will have to teach libtool how to call expr(1) differently on -stable and -current and it looks like yet again different from the rest of the world. :-((( This should exactly match the behavior of any certified UNIX system. Well libtool is still broken, so maybe if systems like that do exist, they don't need libtool? :-))) Yes, I did read the commit message, but I still think the behaviour of the new expr(1) is wrong. Not according to the Standard, or the response from Garrett's request for clarification of the Standard. H. I can understand the requirement to eat '--', but to throw a tantrum just because the commandline started with a '-' is a little too much. BTW, was the response posted somewhere? I searched through -standards, -commit and -current but couldn't find it. Maybe I just didn't ask the right question to the search engine or maybe it was in another list. John -- John Hay -- [EMAIL PROTECTED] / [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: new expr(1) behaviour breaks libtool
On Sun, Apr 21, 2002 at 05:58:22PM +0200, John Hay wrote: Hi, I see the new new behaviour of expr(1) requires you to add '--' if your commandline arguments might start with a '-'. This does break things a little because our old expr(1) does not understand a '--' in the beginning and the new one don't work right without it. :-((( I have a patch for this, but unfortunately it's on a HD I can't access right now. I need to test it on the cluster soon and commit it. Kris msg37498/pgp0.pgp Description: PGP signature