Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Doug Bissett

Pages: 1 ... 64 65 [66] 67 68 ... 107
976
Applications / Re: Apache OpenOffice 4.1.2 for eComStation and OS/2
« on: December 28, 2015, 07:24:19 pm »
Quote
Dynamic linking is the way to go and that's where we are already now!

Well, dynamic linking causes more DLLs to be loaded in shared memory space, which is very limited. I am at the point where I need to stop using very useful programs to accommodate that. If something can be done to fix the problems with loading in high shared memory space, we will get some relief, for a while, but eventually we will run into the same problem, again.

Something needs to be done, and soon, or OS/2 will become unusable, again.

977
Applications / Re: Apache OpenOffice 4.1.2 for eComStation and OS/2
« on: December 27, 2015, 10:04:59 pm »
David has it right. The configuration that you want to use will depend a lot on how your various programs use shared memory space. You also need the 106 kernel from eCS 2.2 b2, which has a couple of fixes for loading and unloading, DLLs in high shared memory space. Unfortunately, it seems that there is still a problem with that. Somebody is working on it, but there are no results, yet.

There are a few things that a user can do to help:

Add the line
dllbasing=off
near the top of CONFIG.SYS. This lets the system make more efficient use of shared memory space, at the expense of a very slight reduction in performance (you might notice it if you are running on a 486-SX processor).

Close, and/or don't use, unnecessary programs.

If you use X(e)Center, remove any DLLs that you don't use. Years ago, I found that WinList kept on using more, and more, shared memory space, even if I didn't use it. That may have been fixed, but I find that LSwitcher works much better.

Keep the number of tabs in Firefox (or Seamonkey) to a minimum. Each one adds a little more.

You can try adjusting VIRTUALADDRESSLIMIT=xxxx in CONFIG.SYS. Some say that 1024 is a good number. Others say that 2048 is good. The collected wisdom says that 1536 is generally the best. Personally, I find that 1536 works best, for me, and I start to have more problems if I increase, or decrease it. YMMV. Unfortunately, I still have problems from both sides, if I don't try to keep usage under control.

Now, for the big one (IMO). Programmers need to start using more static linking, to eliminate the use of DLLs. For some reason, they seem to be very reluctant to do that, but huge packages, like AOO, and Firefox, are eating up large amounts of shared memory space. Using high shared memory space is a partial fix, but it still has problems (the system can crash when the DLLs are unloaded). Running large programs, that use a lot of shared memory space is not a problem when the user only runs one program, but it becomes a large problem when somebody wants to run AOO and Seamonkey at the same time, along with a dozen other programs. It isn't possible to modify a lot of the older programs to use high shared memory, or even stop them from using DLLs (we don't have the source). Other programs, like the CLAM Antivirus daemon (CLAMD.EXE), use large amounts of shared memory space. The older versions used low shared memory space, while the latest (only available with RPM/YUM AFAIK) seems to be using upper shared memory space. That change makes it usable again, but there is still a limited amount of upper shared memory space available. The other option is to use it without the daemon. That is okay, but then it takes about 1 minute to load the database, before it can run the scan, which can take less than a second.

Some of these problems will be slightly easier to manage, when Arca Noae gets their Package Manager (ANPM - previously called Yumie) released. The program itself is working great, but getting it properly installed is turning into a nightmare because of the complexity of YUM itself, and even worse, because users have tried to outsmart the existing systems, or RPM/YUM has not been properly used in the past. RPM/YUM is a terrible choice for managing packages, but that is what we are presented with. ANPM is a user friendly front end for YUM, but it still uses YUM in the background, so it depends on YUM doing the right thing (and all too often, it screws up). It will probably take a bit of work to get it all sorted out, but managing the whole thing should be easier in the end. It will be even better, if the WarpIn interface gets added so it can manage both install methods.

978
Applications / Re: Apache OpenOffice 4.1.2 for eComStation and OS/2
« on: December 27, 2015, 06:09:42 pm »
I also installed over 4.1.1, with no problems. I did NOT use the RPM package which still puts the RPM supplied DLLs in the program directory. That will REALLY screw things up. They should be in a common directory (\eCS\DLL, or \os2\DLL) because other programs use them, and you will end up with duplicates, and down level files, eventually. You won't have the problem if you use RPM/YUM to get those files because it puts them in the YUM directory structure (which has it's own problems).

I also did the procedure to load the DLLs and code high. I cannot even load the program, when Firefox is running, without doing that. I tried loading the Firefox stuff high, but it was crashing (the whole system) about 3 out of 5 times that I closed it (yes, I do have the 106 kernel). So far, AOO4.1.2 hasn't crashed on close, but 4.1.1 did do it about 1 in 10.

Sure hope somebody can fix some of those shared memory problems. My system is running on the hairy edge of disaster, and all I can do about it is stop running some very useful programs.  :-\

979
Storage / Re: Share a 2TB USB drive between Windows and OS/2?
« on: December 19, 2015, 11:06:22 pm »
Quote
2T, that's a nightmare, what do you do with it?

Hehe, now that is a good question. There was an old saying that goes something like: "Give them the resources, and they will find a way to use them." More often than not, it is all too true. The biggest problem is that users tend to collect so much JUNK, that they can't find what they thought they stored there anyway.

Quote
However, as I am fairly sure Doug and others will agree, only 1 disk is not the way to go.

Well, that depends a lot on how you use the machine, but yes more than one, PHYSICAL, backup is generally a good idea. Of course, under normal circumstances, the backup for your backup is the original. However, when Murphy sticks his nose into the mix, all bets are off (sounds like he got you good).

It is always a good idea to have an emergency backup. Mine consists of a USB stick, where I store all of my own personal data, and a few other things that would likely be difficult to get elsewhere. You can always get the OS, and most  programs, from other sources, but you can't get your own address book, or e-mail, from other sources.

FWIW, I don't recommend storing more than one copy of a backup (including incremental, and/or differential backups, if you do it that way) on a single backup media. If it goes bad, you have lost the works. It is also a good idea to store an emergency backup off site. 14 backups, on 14 disks, won't do you much good if they are all destroyed by fire (for instance).

One of the newest backup media, is the "cloud". That might sound like a good idea, but in most cases, it is way too slow  for doing normal backups. You also run a higher risk that somebody will break into it, and do bad things with it. My personal view is that the "cloud" is useful for sales people who need immediate access to things like price lists, and order forms, when they are out of the office. It is not a good place to put  backups, or even photos.

980
Storage / Re: Share a 2TB USB drive between Windows and OS/2?
« on: December 18, 2015, 08:39:27 pm »
I don't know of any OS/2 backup program that would really supports external USB hard drives. Ok, it's possible to copy files to them, but how long would a 500 GB or 1 TB backup run in real world. Good chances that it will not complete at all because the lack of reliable USB device drivers under OS/2. Now try to a restore to bare metal from this external drive, quite impossible in an OS/2-only scenario. External SATA or Firewire-800 connections are much faster than USB 2.0/3.0 and have lower cpu load btw.

What are you talking about? In the real world, USB with 500, or 1 TB (and probably 2 TB - I don't have one) disks work very well (slow, but well). I do multi gigabyte backups to them often, using eCS and the latest Arca Noae drivers (or Lars' drivers). Restore works just as well. I use ZIP to make the backups, and UNZIP to do the restore. I also use RSync to make backups that are much easier to extract individual files from. I do use JFS formatted drives for OS/2 (eCS) backups.

I do agree that eSATA is faster. I don't know much about Firewire, so I can't comment, other than we don't have a working driver. I also agree, that USB does have a heavy CPU load, but that is the way that it was designed, and that is true for all platforms. None of that makes USB drivers "unreliable" for backups. Restore has always worked, for me, when I use a program that is capable of doing it (BackAgain/2000 can't do it any more). Sometimes, SYSINSTX is required to fix the boot record. IMO USB hard disks are far more reliable than using CD or DVD media, and they are far more reliable that tape could ever be (I won't comment on diskettes).

One thing, that I do know, is that using the FAT32 cache program, with a FAT32 formatted drive, WILL cause unreliable operation, no matter what the media is, or how it is attached. It also slows the whole thing, and probably increases the CPU load. I highly recommend REMing the FAT32 cache program in CONFIG.SYS, if you use it. I also recommend not using EAs with FAT32 (very few things, other than the desktop, need EAs). Of course, I don't know why anybody would want to back up OS/2 (eCS) to a FAT32 drive anyway, but stranger things have been known to happen.

981
Storage / Re: Share a 2TB USB drive between Windows and OS/2?
« on: December 18, 2015, 02:06:28 am »
I don't have a 2 TB drive, but I did split a couple of 1 TB drives to share between OS/2 and windows. I used DFSEE to wipe the front end of the drive, then I used Manage Volumes (MiniLVM) to build the volumes. The one for windows was not assigned a LVM drive letter (select 'none'). Then, I used OS/2 to format the OS/2 volume (JFS), and I used windows to format the windows partition (NTFS). If you want to share a volume between OS/2 and windows, I suggest a FAT32 volume, formatted using the DFSEE routine.

982
Applications / Re: What is the best CD Based imaging software
« on: December 15, 2015, 08:39:14 pm »
DFSEE can do whole volume, or whole disk, images, but they need to be restored as a complete package, you cannot pick individual files or directories (Acronis would be the same, when backing up OS/2 volumes). The other problem is that it will restore to the same size, and geometry, not allowing changes, so you end up with exactly what was backed up, losing any interim changes. I highly recommend using DFSEE to do the Disk Analysis thing, and save the results (as prompted).

Since BackAgain/2(000) has become totally unreliable (it backs up okay, but it won't restore properly, in all cases), I have been using ZIP to zip up whole volumes (including active boot drives). It is easy to simply format a volume, then unzip into it (size and geometry don't matter, as long as it is OS/2 compatible). Sometimes, I need to use SYSINSTX, but it has always worked, for me. I use that method for boot drives, but I do make sure that they contain ONLY the OS. and a few other things that don't matter. I also ZIP up my programs/data drives, but that is a secondary backup. You do need a version of ZIP/UNZIP that can handle files that are larger than 2 GB, if you have files larger than 2 GB, or if the resulting zip file is larger than 2 GB. Note also that this is NOT an incremental, or differential backup solution.

For my data/programs volumes, I also use RSync to copy to a USB drive, and a LAN drive. To extract individual files, or directories, I just need to attach the device, and copy what I want. USB, and LAN, is slow, so if I ever need to copy a whole volume, I would probably remove the drive from the USB enclosure, or the LAN machine, and attach it directly to the computer. This is also not really an incremental, or differential, backup solution. It always produces a full backup by replacing, adding, or deleting, files to match the source. After the first pass (where it has to do all of it), it simply does the changes, so it is very fast.

I do use the Seagate version of Acronis (free, but you do need a Seagate, Maxtor, or Samsung drive attached - can be USB, IDE, SATA, SCSI, SAS, but not LAN) for windows backups (Look for Seagate Disk Wizard - it does not have all of the features of the full Acronis program, but it works well). There is no doubt that that is the very best windows backup (the ONLY one that actually works for windows, that I know about), but it is probably not ideal for OS/2 (if it works at all). Be sure that you don't let it mess with boot records, or it could mess them up.

983
Utilities / Re: OS/2 arrows: 2.1: What components to include to OS?
« on: December 14, 2015, 04:12:17 am »
Quote
so I will probably ask about hot swap in that very related bug. The reason the drive was attached at boot was simply because hot swap does not work

Don't report more than one problem per incident, or both problems may be ignored. Hot swap is completely different from your "unable to operate" problem. It is  also not a deadly problem because a simple boot works around it.

Quote
The problem does not exist with the slightly earlier, slightly smaller 1.32 supplied with the eCS2.2beta2 release.

The AHCI driver that comes with eCS 2.2 b2 does have some sort of problem, it is best not to use it, although my understanding is that it either works, or it doesn't. There is a later version, also called 1.32, that was made available after eCS 2.2 b2 was released. It fixes the problem, and it would be very interesting to see if it also has the problem that you see with the Arca Noae driver. If so, it was that change that messed it up. If not, it was whatever Arca Noae did to it. Unfortunately, the eCS betazone seems to be gone, and I think that was where the fixed 1.32 driver was.

Quote
The possibly good news is that eSATA is simply SATA but with hot swap capability

The SATA hardware is hot swap capable. ESATA simply adds a defined cable to do it. I suspect that the only real difference is extra shielding, for longer distances, and possibly a tougher connector.

Quote
I also suspect the driver needs to be "hot swap" capable and have no idea if that is the case with the current/recent os2ahci.add releases.

The "removable" feature needs to be added, otherwise it would work. The other thing that it needs, is a hook to allow an eject, which needs to close out the file system, before telling the user that they can disconnect it. I don't think that ACPI needs to get involved any more than it gets involved in ejecting a USB drive. There is no need to power off the drive before connecting/disconnecting it.

984
Article Discussions / Re: [Arca Noae] Checked your disk lately?
« on: December 14, 2015, 03:38:29 am »
Quote
do this setting got reenabled for you when you install a new version? Then it's a bug of course. Or do you make new installations so often that this is a problem for you?

No, it doesn't change for an update to x(e)WP. I have been doing a lot of new VBox installs lately, while testing things. I usually don't think about it until I shut down, then, I forget again after I start up. It is not really a problem, but it is annoying until I make a specific boot, just to get it turned off. I realize that some new version of OS/2 will be required to get an updated XWP into it, but if it starts now, it will get there eventually.

985
Article Discussions / Re: [Arca Noae] Checked your disk lately?
« on: December 12, 2015, 01:33:16 am »
Quote
To my understanding that was the reason why SHL added the delay in xwp code. It's another story that this version of xwp source code never found the way to the public until I started uploading betas.

It sad that there seems to be no mention of what was changed, or why. Actually, that isn't entirely true, there is a list of things that changed, but the average user (including me) has no idea what it is talking about. Whether 200 ms is enough delay remains to be seen. I am not willing to change my main system, yet, but I will change other, less critical, systems (including the one where I had the problem, but it has a different hard disk now), and see what happens.

In Pete's case, he has a problem that sounds like the one that I had, and adding the delay fixed mine, so he should try it. If it fixes it, and he meets the rest of your conditions, I would suggest a longer delay in XWP. He didn't mention which version of XWP, or eWP, he is using, so the delay may work. In any case, I would suggest that the delay be left in the code, for diagnostic reasons, but it could probably be less than 4 seconds.

Quote
This all is a workaround for a problem either not understand or not solved in the right way.

There are probably 100 (or more) of those in eCS. In a lot of cases, the work around is necessary because there is no way to fix the problems properly.  In other cases, nobody is interested in doing it. And in even more cases, those who might be able to fix them, insist that the problem doesn't exist. I suspect that most of the work arounds come from somebody who can sort it out, who fixed it because they had the problem. The rest of us are left hanging, and anything that somebody can figure out becomes the work around.

While I have your attention, why don't you turn off that annoying animation that is default on XWP shutdown? At least make it so that it is not the default.

986
Article Discussions / Re: [Arca Noae] Checked your disk lately?
« on: December 09, 2015, 06:45:56 pm »
Quote
Honestly I do not know how long the xwp delay will be when the button is checked. Is it configurable?

It is 4 seconds, and produces 4 beeps when it happens.

Quote
If there is no reason other than a workaround for other drivers fault, then this has to be discussed and solved in that driver.

I don't think it is the driver that was causing the problem that I had, a few years ago (and I suspect that this might be Pete's problem). It seemed that the driver thought all disk activity was done, but the drive still had some stuff in it's internal cache when the machine finished processing the shutdown, and cut the power. The delay fixed that problem, and 4 seconds at shutdown is hardly noticeable. It probably doesn't need to be more than 1 second, but machines are getting faster, and disk caches are getting bigger, so some delay is probably a good thing. Why not use more time than anybody might think is necessary. 200 ms might be enough, most of the time, but so is 1000 ms. 199 ms may not be enough. Who really cares how long it takes to shut down a machine, as long as it gets done right. This is not a race. Well, I suppose it is a race between getting all of the data written, and pulling the plug. Getting all of the data written is much more important than pulling the plug.

I will note, that I have always had to use /FS on the USB drivers, in that same machine, or the BIOS, very often (3 out of 5), would hang on initializing the USB stuff on a warm boot. It is very rare to have any USB device attached at shutdown (except for the mouse, and the UPS). I never really saw any other problems with USB. BIOS still has a problem initializing the USB, but it is about 1 in 100 with /FS. I have used /FS on other machines, and it doesn't make any difference, so I would suggest that /FS should be the default (why not?).

987
Article Discussions / Re: [Arca Noae] Checked your disk lately?
« on: December 08, 2015, 06:15:37 pm »
Quote
There should be no additional delay necessary with current xwp and current ACPI.

Some years ago now, I did have a problem with a new machine, where it was getting the power turned off before the disk finished writing something. I didn't see any reason not to use the delay (obviously it was implemented for some reason), so I just started using it always. I still don't see any reason not to use the delay, it is definitely safer than not using it. Note that turning off the power, before the disk has finished writing, is not the same as not being able to load the power off module because the file system had been shut down. It sounds like Peter's problem is that the power is going off before the disk is finished, so the delay may be necessary (worth trying anyway).

I will note, that not all machines use ACPI. My ancient ThinkPad A22e doesn't use it, and it looks like it might need the delay (but I haven't tried without it, for a long time). I will also note, that that machine does suspend/resume, with no problem, and has done so since Intel fixed their network driver, oh so many years ago. I have yet to get a modern machine to do that.

While I am at it, I have one old machine that will not power off using ACPI, or APM. It goes through the routine, and something clicks when it should go off, but it just does a power on again. Of course windows will power it off, and even using DEL when Air Boot opens, will power it off. The delay makes no difference. I will note, that windows XP refuses to use ACPI, but eCS uses it with no problems, other than it won't complete a power off. Very strange...

988
Article Discussions / Re: [Arca Noae] Checked your disk lately?
« on: December 07, 2015, 04:05:10 am »
Quote
I tend to suspect the possibility that JFS 1.09.06 does not get shutdown properly sometimes during system shutdown or reboot although all seems to work properly.

This is probably the reason why there is a shutdown option to delay the actual power off. It has been in eCS for quite a while, I think it is really part of X(e)Workplace. Desktop properties-> X(e) Shutdown, near the bottom. It never hurts to use it, but it may hurt if you don't, when you need it (It is definitely a possibility). I am not sure why it is optional, it should always do the delay, to let the system finish what needs to be done before the power is cut.

989
Mail-News / Re: PMMail 3.11.30.1767 and PGP (PGPTools 1.01)
« on: December 01, 2015, 10:21:34 pm »
PGPTools, is a front end for PGP. You do need to install PGP to use it. PMMail will use PGP, when it is properly installed according to the instructions in the PMMail help. PMMail does NOT use PGPTools.

You should also consider updating to PMMail 3.20: http://pmmail.os2voice.org/index.php?title=PMMail_for_OS/2, but nothing was changed for PGP support.

It may also be a good idea to subscribe to the PMMail user list: http://www.os2voice.org/mailinglists.html, where you can ask questions to those who use the program.

990
Quote
and you said that you did the zip thing a couple of times?

Yes, at least half a dozen times. Most of them were using a ZIP file made from the active boot drive. Of course, a ZIP file from an inactive drive also works. I will note, that I do NOT allow normal software to be on the boot drive. The simple procedure is: boot to an alternate boot system (usually my previous eCS install), format the drive, and unzip the backup to it (I use the eCS Archive Viewer, for simplicity). Reboot, and make sure that it actually works (hasn't failed yet, for me). It may be necessary to use Sysinstx, under some conditions, but I haven't needed it (I only use JFS drives).

I also ZIP up my programs/data drive, as an alternate backup to what I do with RSync. I did do a test restore on one of my test systems, and it worked, but I haven't had to do it on a "live" system, yet (that ZIP file is a little over 12 GB, and it contains a 10 GB virtual disk file for VBox).

It has been about 5 years since I actually tried to restore a boot drive from a BA/2 backup. It didn't work (I was able to reproduce the problem, every time I tried it, on multiple systems), so I went looking for a better solution, and came up with ZIP. I do recommend the latest version, so it can handle files (in and out) that are larger than 2 GB. Boot drives probably don't need that, but it doesn't hurt. I also looked at ARJ, and RAR, but they were slow, and clumsy, compared to ZIP (ARJ seems to be better, IMO, but it has way too many parameters, most of which are completely useless). What ZIP/UNZIP needs is a front end so the user can do easy selection of files and directories, for backup and restore.

Quote
And finally as for Rsync it chokes in a lot of situations

Actually, it doesn't. I have reported a crash problem when there are invalid EAs on a file, and SHL says he will find a way to fix the problem (apparently it is something in the kernel that doesn't work right). ZIP has trouble with those files too. I still haven't figured out where the invalid EAs are coming from (I suspect EVFSGUI, but I can't prove it). Other than that, I haven't seen any problems with RSync, for a LONG time, and I use it a lot.

Pages: 1 ... 64 65 [66] 67 68 ... 107