At 09:58 p.m. 3/11/2003, you wrote:
Hello,
This is the second Redhat server I have used for our clients and I am quite frustrated with it. The other one is Redhat 7.2 which has been (relatively) okay. I am so scared by this experience with Redhat 9.0 (and not meeting my client's expectations) that I am considering to 'cut my losses' and switch to FreeBSD, Solaris, or (maybe) RH 7.2.
In short: I cannot get much of anything to compile without warnings, errors, and configure scripts failing to work. Here is just a short list, none of which occur on my FreeBSD, RH 7.2, or Solaris systems:
#1. qmail-1.03 - won't compile, "undefined reference to errno"
There is a patch on the qmail site for this error - its actually quite simple to fix and came about due to mistakes in programming techniques on the program authors part - The way Qmail (and from the looks, PHP) reference errno (a C programming thing) was not "standard" (which was probably necessary for multi-platform support at a wild guess) and so when the GCC guys decided to start standardizing things broke.
If you cant find the patch then let me know and I'll hunt it down and flick you a copy. If you are installing qmail tho you may want to look into the qmail toaster for redhat (I think it supports 9) and includes a _lot_ of really good stuff (anti-spam support, vpopmail, web administration and most of the "standard - should be put together to make qmail 1.04" patches) - again, if you cant find the link then let me know and I'll flick it to you off list.
#2. php-4.3.1 - won't compile when --with-mysql is specified "undefined reference to errno" (presumably this is fixed in the CVS version of PHP, but I cannot really consider running a CVS version of anything in a production environment)
Try a newer version, or it could be the version of the libmysql it is trying to link against (or the built in support if you don't have the mysql-dev stuff installed) This will be the same error above, if you get really stuck, again let me know and I may be able to help.
#3. php-4.3.1 - when --with-mysql=/usr is specified (causing PHP to not use it's internal mysql code), I get the following errors: A. warning: implicit declaration of function `getpgid' B. warning: implicit declaration of function `pread' C. warning: implicit declaration of function `pwrite' D. the use of `tmpnam' is dangerous, better use `mkstemp`
#4. apache-1.3.29 - the use of `tmpnam' is dangerous, better use `mkstemp
This sounds more like an apache error than redhat, keep in mind, faults with programming techniques get found all the time as with other bugs, and in this case you may find a package such as autoconf has detected a "bad form of programming, try this instead" type error. The same goes for the errors reported above, if "make" finished without a fatal error then you can pretty much bet everything will work ok - its just debugging stuff for the people that like to compile their own code.
#5. htdig-3.1.6 - configure fails 'checking for fstream.h'... (yes the
g++ rpm is installed).. Making me wonder:
A. Are configure scripts for other things failing to detect header
files and 'falling back' to some deprecated APIs?
B. Could that be what the tmpname error is related to?
C. How can I possibly rely upon this development environment?
A) Probably not, but you'd possibly have to read the configure script and log outputs to see for yourself, this is why many people like to use rpm's as they are pre-built and all this programmy stuff doesn't give them a headache. Check that you have the latest version of htdig.
B) probably not. The tmpname error relates to "easily guess able names for temp files" as this could cause some security specific attacks in some instances and the programmer should be aware that there may be a huge .000001% chance that some nasty person may be able to exploit this !" What you will also find, is that in the year 2013 mkstemp will be found to have a similar hole in it allowing a nasty attacker to also guess temp filesnames with a wopping 0.00000004% chance and you'll get a warning if a program is compiled with that function :-) This error is a warning only - and only relates if the program uses the "temp file creation" function to create a temp file that may contain confidential data - in the case of apache this is probably not the case.
... As I am sure anyone with a large number of servers to manage is aware, having odd development environments which require 'special handling' is a nightmare to deal with. It blows my mind that I have only started to configure this server and so far 5 out 5 software builds have 'failed' (by my standards) to configure, compile, or install properly.
It actually surprises me that you are using compiled packages rather than the standard RPM's - updating your systems can be a lot easier when you use pre built binaries or compile your own packages and install those. Usually if you look into the reasons for the errors in compiling you will find someone else has come across the same problem and will usually explain why it happened, over time you get to know what are issues with the OS upgrades and what are issues with the packages themselves, in this case, the problems are more "new version of the C Libraries" than a development OS.. believe me, BSD will have the same problems as will most other OS's (Solaris has its own quite nice set of issues you'll need to deal with :-) )
[snip]
My specific questions are: 1. Am I doing something wrong trying to configure/compile this software?
Not really - although I would probably no longer do things that way unless really needed. (I used to :-) )
2. Should I consider the installation of RH 9 to be suspect at this point?
Nope.
3. Is there something wrong with the RH 9 development environment?
Nope - Unless you are unfamiliar with development environments :-)
4. If I stick with RH Linux, should I downgrade to 7.2?
If you did consider a downgrade, I would strongly suggest RedHat 7.3 if you can get it.
5. What are your thoughts on RH's long-term strategy in this area? Is RH 10 going to introduce the same magnitude of problems all over again? Should integrators/administrators anticipate these sort of problems (which don't seem to exist to this level in other OSes) when upgrading?
RedHat 10 doth not exist - Read back on the threads on Fedora if you are interested, and believe me, these same problems exist and in some cases can be even more tricky to solve.
Sorry if I seem overly bitter - I've been fighting with this all weekend and have now missed a very important deadline as a result. These packages I am trying to compile comprise about 10% of the software I need to install on our servers and if I can't even get through these I can only imagine what lies in store for me as I move on to the more complex ones which are not updated as frequently....
:-)
If you have problems with those then feel free to post here as well - someone should be able to help you out.
As for your bitterness, think nothing of it - we all have bad days..
-- Steve.
-- redhat-list mailing list unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe https://www.redhat.com/mailman/listinfo/redhat-list
