How did bsdinstall get released?

I am just running into bug after bug in the new installer. I've already documented some of them in this forum, submitted one bug report, and there are others. For example, if you create a gpt partition table, manually create partitions and later have something go wrong with the install and try to reuse the existing partition table (adding back the mountpoints for each partition), you will invariably get an error in the subsequent install:

Code:
Error while extracting base.txz: Can't set user=0/group=0 for var/emptyCan't update time for var/empty

I have said before and I will say it again that there is something wrong with FreeBSD's QA process. Gross bugs that never should have seen the light of day in a properly tested system find their way into FreeBSD releases.

This message will bring out the FreeBSD apologists. It has in the past when I've said similar things. Have at it, guys. But I'm done. I've been burned by this system too many times; I keep trying new releases, hoping against hope that things have changed. And they haven't. I'll do my computing with something that works -- Linux, OpenBSD.
 
bsdinstall is still kind of a technical preview. But something just had to replace sysinstall. If you need to create an unusual setup, you can do it via CLI.
 
pkubaj said:
bsdinstall is still kind of a technical preview. But something just had to replace sysinstall. If you need to create an unusual setup, you can do it via CLI.

It's the installer in 9.0-RELEASE! If it's known to be beta-quality software (I don't think it's even at that level), then it should be made available as such. To have made it the installer in a release in the condition it's in is just baffling to me.
 
There are other options. PC-BSD's installer will install plain FreeBSD.

Background: sysinstall(8) is reputed to be a mess. pc-sysinstall(8), the PC-BSD installer, was going to become FreeBSD's installer. However, it had problems installing non-i386 architectures. So bsdinstall(8) was written. It has some capabilities that sysinstall(8) lacked, but is missing a lot, too. There has been a lot of noise and complaints, with little actual progress. I would like to see pc-sysinstall(8) become the default. How to make that happen? Getting involved with PC-BSD, the FreeBSD mailing lists, or the FreeBSD Foundation are probably most effective.
 
gordon@ said:
Patches to fix bugs in the software are happily accepted.

I have already suggested two changes to the installer documentation, where the guide was not consistent with the software and have submitted a PR relating to specification of static IP addresses not working correctly in the installer, at least under certain circumstances.

I'm happy to submit patches, but I did not install a release of FreeBSD in order to become a FreeBSD developer. I have work that I want to get done with this system. But having gotten over the snit brought about by tripping over the nth bug in the installer, if and when time permits, I will look over the code to see if I can fix some of these things (since I'm aware of how to reproduce the problems), though I'm not using the system and no longer have plans to do so.

But my overriding concern is the thinking that led to the release of this installer. Just ask yourselves whether you think Theo de Raadt would have signed off on this. I've never seen anything this broken in either OpenBSD or in a good Linux distribution, such as Slackware. And the previous post that said "But something just had to replace sysinstall" is missing a word. Something *better* was needed to replace sysinstall and the current situation is a big step backwards in my opinion. It has implementation bugs and design bugs (Since / needs to be the first UFS partition, why no ability to set the size of a partition to '*", indicating use all the available space remaining after the other partitions have been allocated, ala the NetBSD installer?).
 
The bsdinstaller was tested on non-standard ports of PPC/PPC64 along with SPARC, i386, amd64. In the installation guide for PPC 32 bit, there are a few references to the guide.

Some notes that may not be included:

On PPC and PPC64 systems, you may want to use a Linux Disc to create more than four partitions.

I haven't installed in a few months due to personal matters; however, the past few times required use of a Linux disc.

The bsdinstaller seems only to have options for FreeBSD-UFS(2) and FreeBSD-swap. You will still need to use sysinstall() for other partition types. Again, since it has been a few months, this information may have changed.

Gpart gpart()- Since capitalization of the item prevents the manual page from being accessed, I had to add the man reference. This explains the double entry.- on sysinstall() is much cleaner in reference to code than the Linux counterpart. Nathan Whitehorn and others- Hibbits, Dmitri, et al- told me not to worry about trying to clean up the Linux gpart code.

Gpart gpart() can be used to manipulate previously created partitions that you want to dedicate to FreeBSD.

Besides the use of sysinstall(), one can use fdisk().
 
donallen said:
(Since / needs to be the first UFS partition, why no ability to set the size of a partition to '*", indicating use all the available space remaining after the other partitions have been allocated, ala the NetBSD installer?).
You need to manually set the partition size when using gpart() unless you have previously created partitions with a Linux disk. If such is true, then using gpart() will allow you to change the partition type.
 
donallen said:
I keep trying new releases, hoping against hope that things have changed. And they haven't. I'll do my computing with something that works -- Linux, OpenBSD.

Assuming you've never been happy with sysinstall either, then you'd probably be better off sticking to Linux / OpenBSD. FreeBSD hasn't changed that much over the years, and I really doubt any new features will be that drastic as to convince you otherwise.

I've installed a lot of operating systems over the years. I've had problems with every Microsoft installer since Windows 3.1 (exempting Windows ME and Win2003). Mac OS installers seem to always work but spit out a lot of error messages. I've got a long list of Linux installers that never worked quite right either. bsdinstall is among the buggiest I've used, but I've been able to get through it. It's just one install though, and usually years of use after that. It sounds like you're installing more than using, so probably not a good fit as far as the OS for you.
 
It's yet another case of a problem in software testing, who is going to test it and report the bugs? The people who are capable of testing it and can provide good feedback don't really need it because they can install FreeBSD without using bsdinstall(8) . There were lots of calls for testing bsdinstall on the mailing lists before FreeBSD 9 was released but apparently not many bothered.

On the other hand the people who the tool is aimed at are usually incapable providing useful bug reports because the reports basically just read "it doesn't work when I do this" with no context and background information.
 
Morte said:
Assuming you've never been happy with sysinstall either, then you'd probably be better off sticking to Linux / OpenBSD. FreeBSD hasn't changed that much over the years, and I really doubt any new features will be that drastic as to convince you otherwise.

I was never particularly unhappy with sysinstall. It certainly wasn't great, but it worked, for the most part. And as for the system, I'm not looking for new features; I'm looking for improved quality. And there has been some of that over the years. For example, the USB layer before the 8.x releases was a disaster. It got re-written and seems to be solid now.

Morte said:
I've installed a lot of operating systems over the years. I've had problems with every Microsoft installer since Windows 3.1 (exempting Windows ME and Win2003). Mac OS installers seem to always work but spit out a lot of error messages. I've got a long list of Linux installers that never worked quite right either. bsdinstall is among the buggiest I've used, but I've been able to get through it. It's just one install though, and usually years of use after that. It sounds like you're installing more than using, so probably not a good fit as far as the OS for you.

Among the installers that work well are those of NetBSD and OpenBSD, both smaller projects than FreeBSD. The Slackware installer works beautifully, in my experience, as does the Arch installer. Debian's installer is quite good also. There are plenty of examples of installers dealing with problems similar to that of bsdinstall that are of far greater quality.

Your point about installing once and then having years of use is bogus for a number of reasons. FreeBSD updates are released with a periodicity less than "years". To take advantage of them, you need to run the installer. Furthermore, having the first thing you encounter in release 9.0, the installer, be such a mess is hardly an incentive to invest more time in this system. What else has been included in this release in this condition? And my past experience with FreeBSD has always turned up something of this sort that made the system unusable for me. Yes, you are right that FreeBSD is not a good fit, but for the reasons I've given, not the reasons you gave.
 
kpa said:
It's yet another case of a problem in software testing, who is going to test it and report the bugs? The people who are capable of testing it and can provide good feedback don't really need it because they can install FreeBSD without using bsdinstall(8) . There were lots of calls for testing bsdinstall on the mailing lists before FreeBSD 9 was released but apparently not many bothered.

On the other hand the people who the tool is aimed at are usually incapable providing useful bug reports because the reports basically just read "it doesn't work when I do this" with no context and background information.

If it wasn't properly tested, it shouldn't have been released. And complaining that you couldn't get people to test bsdinstall begs the question of how NetBSD and OpenBSD managed to produce excellent installers, projects with fewer resources than FreeBSD? How does Patrick Volkerding do it with Slackware, another project not exactly flush with money or people?
 
donallen said:
FreeBSD updates are released with a periodicity less than "years". To take advantage of them, you need to run the installer.

That's not correct. You can upgrade without going near an installer. Cvsup is one option, freebsd-update another.

There was a comment above about capable testers not needing the bsdinstall to deploy the OS. I'm not sure how you would be able to deploy FreeBSD on a new system without using bsdinstall, though I don't doubt it's possible.
 
It's terribly simple to install FreeBSD without bsdinstall or other automated installer once you understand what the automated installer does. I won't go to details here but it basically involves partitioning the disk, creating the filesystems and extracting the needed distribution tar(1) archive files on the filesystems. Add a few bits of initial configuration and there you have very minimal but yet fully functional FreeBSD installation.

I do admit the above is not the way a UNIX newbie would be able to install the system but anyone with a little bit more experience can do it.
 
Jimmy said:
That's not correct. You can upgrade without going near an installer. Cvsup is one option, freebsd-update another.

I stand corrected on this point, but it doesn't change my conclusion about the use of FreeBSD.
 
kpa said:
I do admit the above is not the way a UNIX newbie would be able to install the system but anyone with a little bit more experience can do it.

The problem is, that the first impression one gets from their experimentation with a new OS is via the installer. If the installer is a bug-ridden pile of garbage, then they may simply give up and move onto something more reliable/well tested.

I hear sysinstall was a clunky old thing, but at least it worked.

bsdinstall prompted me to learn how to install manually. A noob won't bother.
 
throAU said:
The problem is, that the first impression one gets from their experimentation with a new OS is via the installer. If the installer is a bug-ridden pile of garbage, then they may simply give up and move onto something more reliable/well tested.

I hear sysinstall was a clunky old thing, but at least it worked.

bsdinstall prompted me to learn how to install manually. A noob won't bother.

And some experienced people won't bother, either. Your point about the installer being the first thing one sees is exactly right and leads to the conclusion I came to: if the installer made it into the latest release in the condition it's in, the QA/release process must be broken, and so this system isn't worth a further investment of time and effort. As you say, there are other reliable, well-tested options, e.g., Slackware, Debian, OpenBSD.
 
Back
Top