I wish and hope there was edit and continue in windbg. I don’t know how
incrementally complex it would be to support that. But I would guess,
not very difficult, since all the pieces seem to be there in some form
or the other (from MS perspective) already.
Time to put them all together perhaps in the next major release of
windbg/wdk ?
windbg doesn’t support editing of any kind, so you’re asking windbg to
turn into a full IDE. That is a huge change and not something we’re
planning to do.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
Sent: Monday, December 19, 2005 12:06 PM
To: Kernel Debugging Interest List
Subject: [windbg] Edit and continue
I wish and hope there was edit and continue in windbg. I don’t know how
incrementally complex it would be to support that. But I would guess,
not very difficult, since all the pieces seem to be there in some form
or the other (from MS perspective) already.
Time to put them all together perhaps in the next major release of
windbg/wdk ?
You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com
Well I for one don’t.
First, Windbg is primarily a kernel debugger and the fact that I am many
times forced to reboot when working on the drivers upon which I work does
not bother me. I’ve been doing this for over ten years and booting a
system, to me, simply ensures that I have a fresh environment that I can
go and pollute with whatever it is that I am about to pollute it.
Therefore, “Edit and Continue” is wasted on me, and possibly many others.
Second, I would much rather see a fully functional version of Windbg with
few if any bugs. I really would rather that the Windbg development team
NOT waste their time adding bells and whistles to make Windbg the end all
development environment. Simply make it work and make it reliable.
Gary G. Little
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
Sent: Monday, December 19, 2005 2:06 PM
To: Kernel Debugging Interest List
Subject: [windbg] Edit and continue
I wish and hope there was edit and continue in windbg. I don’t know how
incrementally complex it would be to support that. But I would guess,
not very difficult, since all the pieces seem to be there in some form
or the other (from MS perspective) already.
Time to put them all together perhaps in the next major release of
windbg/wdk ?
You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com
Satya Das wrote:
I wish and hope there was edit and continue in windbg. I don’t know how
incrementally complex it would be to support that. But I would guess,
not very difficult, since all the pieces seem to be there in some form
or the other (from MS perspective) already.
Windbg is handled by a very different team from the folks who do Visual
Studio. Edit-and-continue requires a level of cooperation between the
debugger, the compiler, and the linker that just isn’t present in
Windbg, and in my opinion, is not desireable.
Time to put them all together perhaps in the next major release of
windbg/wdk ?
This is one of the very best reasons for learning x86 assembly language,
even in this era of hypergood optimizing compilers. When you know x86
assembly language, you can do the “edit and continue” on your own, by
patching the instructions just before you get there.
One problem with “edit and continue” in a driver is that mistakes in a
driver very often result in the sudden and unfortunate demise of your
Windows session. If the problem is recoverable, it isn’t that much
harder to unload, recompile, and reload.
–
Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.
Agreed. We have edit-and-continue – it’s called .reboot .
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of
xxxxx@seagate.com
Sent: Monday, December 19, 2005 3:21 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
Well I for one don’t.
First, Windbg is primarily a kernel debugger and the fact that I am many
times forced to reboot when working on the drivers upon which I work does
not bother me. I’ve been doing this for over ten years and booting a system,
to me, simply ensures that I have a fresh environment that I can go and
pollute with whatever it is that I am about to pollute it.
Therefore, “Edit and Continue” is wasted on me, and possibly many others.
Second, I would much rather see a fully functional version of Windbg with
few if any bugs. I really would rather that the Windbg development team NOT
waste their time adding bells and whistles to make Windbg the end all
development environment. Simply make it work and make it reliable.
Gary G. Little
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
Sent: Monday, December 19, 2005 2:06 PM
To: Kernel Debugging Interest List
Subject: [windbg] Edit and continue
I wish and hope there was edit and continue in windbg. I don’t know how
incrementally complex it would be to support that. But I would guess, not
very difficult, since all the pieces seem to be there in some form or the
other (from MS perspective) already.
Time to put them all together perhaps in the next major release of
windbg/wdk ?
You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com
You are currently subscribed to windbg as: xxxxx@stonestreetone.com To
unsubscribe send a blank email to xxxxx@lists.osr.com
Throw .kdfiles in the mix and life is generally better (though it doesn’t
appear to support NT4 style drives or export drivers, which is mildly
upsetting).
-scott
–
Scott Noone
Software Engineer
OSR Open Systems Resources, Inc.
http://www.osronline.com
“Arlie Davis” wrote in message
news:xxxxx@windbg…
> Agreed. We have edit-and-continue – it’s called .reboot .
>
>
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of
> xxxxx@seagate.com
> Sent: Monday, December 19, 2005 3:21 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> Well I for one don’t.
>
> First, Windbg is primarily a kernel debugger and the fact that I am many
> times forced to reboot when working on the drivers upon which I work does
> not bother me. I’ve been doing this for over ten years and booting a
> system,
> to me, simply ensures that I have a fresh environment that I can go and
> pollute with whatever it is that I am about to pollute it.
> Therefore, “Edit and Continue” is wasted on me, and possibly many others.
>
> Second, I would much rather see a fully functional version of Windbg with
> few if any bugs. I really would rather that the Windbg development team
> NOT
> waste their time adding bells and whistles to make Windbg the end all
> development environment. Simply make it work and make it reliable.
>
> Gary G. Little
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
> Sent: Monday, December 19, 2005 2:06 PM
> To: Kernel Debugging Interest List
> Subject: [windbg] Edit and continue
>
> I wish and hope there was edit and continue in windbg. I don’t know how
> incrementally complex it would be to support that. But I would guess, not
> very difficult, since all the pieces seem to be there in some form or the
> other (from MS perspective) already.
>
> Time to put them all together perhaps in the next major release of
> windbg/wdk ?
>
>
> —
> You are currently subscribed to windbg as: unknown lmsubst tag argument:
> ‘’
> To unsubscribe send a blank email to xxxxx@lists.osr.com
>
>
> —
> You are currently subscribed to windbg as: xxxxx@stonestreetone.com To
> unsubscribe send a blank email to xxxxx@lists.osr.com
>
>
>
>
I did not mean a full IDE but more like a smarter kdfiles that can
refresh code while debugging.
Thanks
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Drew Bliss
Sent: Monday, December 19, 2005 12:10 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
windbg doesn’t support editing of any kind, so you’re asking windbg to
turn into a full IDE. That is a huge change and not something we’re
planning to do.
There are numerous roadblocks to providing even simple EnC for kernel
drivers. We don’t have any plans to do any kind of EnC support at this
point.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
Sent: Monday, December 19, 2005 2:46 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
I did not mean a full IDE but more like a smarter kdfiles that can
refresh code while debugging.
Thanks
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Drew Bliss
Sent: Monday, December 19, 2005 12:10 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
windbg doesn’t support editing of any kind, so you’re asking windbg to
turn into a full IDE. That is a huge change and not something we’re
planning to do.
You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com
So what stop’s you from doing an Enable/Disable in the Device Manager or
via DevCon, or your own utility to stop/start your driver?
Gary G. Little
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
Sent: Monday, December 19, 2005 4:46 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
I did not mean a full IDE but more like a smarter kdfiles that can
refresh code while debugging.
Thanks
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Drew Bliss
Sent: Monday, December 19, 2005 12:10 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
windbg doesn’t support editing of any kind, so you’re asking windbg to
turn into a full IDE. That is a huge change and not something we’re
planning to do.
You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com
Slightly different use case - where you cant just unload and reload
because of the bug leading to side effect/hang/crash [ you have a typo
or a silly mistake like assigning instead of comparing or you set the
boolean condition wrong in your if check or your dispatch routine
indices are all off by one or your variable was left uninitialized and
you want to initialize before you forget etc ] that becomes apparent
when execution is about to/already hit the code (probably a new change).
Rebooting seems like an overkill for this.
From Drew’s response, sounds like it is too complex to implement a basic
one for drivers.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of
xxxxx@seagate.com
Sent: Monday, December 19, 2005 2:56 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
So what stop’s you from doing an Enable/Disable in the Device Manager or
via DevCon, or your own utility to stop/start your driver?
Gary G. Little
Just think about how you would do something like this. You need to find all
references to that driver. From IoManager and all filters in whatever stack
is involved. Then you have a new driver with code that needs to replace it
while preserving all data changes from the current instance. If the change
was to the size or data order of the device extension, how could it be done?
Doing this to a running OS boggles my mind. It might could be done if
someone had complete control and understanding over all of Windows. Just
thinking about this give me a headache, because I don’t think anyone can
truly understand Windows down to the smallest details in all its subsystems.
DOS - yes. Windows - please shoot me.
It is bad enough when the debugger completely owns the address space of the
target application. There the compiler also has the control over how the
code segments are done so a change can be replaced without impacting areas
outside of that process. I suspect that EnC would have problems with
application that establish system hooks and especially those who load into
other processes.
“Satya Das” wrote in message news:xxxxx@windbg…
I did not mean a full IDE but more like a smarter kdfiles that can
refresh code while debugging.
Thanks
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Drew Bliss
Sent: Monday, December 19, 2005 12:10 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
windbg doesn’t support editing of any kind, so you’re asking windbg to
turn into a full IDE. That is a huge change and not something we’re
planning to do.
Which is a good excuse to learn assembly, as has been pointed out, and patch
it right there in memory to test your theory and then change the source and
rebuild.
Gary G. Little
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
Sent: Monday, December 19, 2005 5:28 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
Slightly different use case - where you cant just unload and reload
because of the bug leading to side effect/hang/crash [ you have a typo
or a silly mistake like assigning instead of comparing or you set the
boolean condition wrong in your if check or your dispatch routine
indices are all off by one or your variable was left uninitialized and
you want to initialize before you forget etc ] that becomes apparent
when execution is about to/already hit the code (probably a new change).
Rebooting seems like an overkill for this.
From Drew’s response, sounds like it is too complex to implement a basic
one for drivers.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of
xxxxx@seagate.com
Sent: Monday, December 19, 2005 2:56 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
So what stop’s you from doing an Enable/Disable in the Device Manager or
via DevCon, or your own utility to stop/start your driver?
Gary G. Little
You are currently subscribed to windbg as: unknown lmsubst tag argument: ‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com
__________ NOD32 1.1328 (20051219) Information __________
This message was checked by NOD32 antivirus system.
http://www.eset.com
Have you noticed that Windbg has a separate disassembly window? What was
wrong with the interlaced source and assembly window? I don’t like to use
the disassembly windows since it is more difficult to keep the source and
binary together. I will use it and can handle doing my machine code
changes, but most of the newer programmers & engineers seem to avoid
assembly. I started with assembly and no one seems to do that any more. It
is all 4GL or Python or Perl or Forth or xxx.
“Gary G. Little” wrote in message news:xxxxx@windbg…
> Which is a good excuse to learn assembly, as has been pointed out, and
> patch
> it right there in memory to test your theory and then change the source
> and
> rebuild.
>
> Gary G. Little
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
> Sent: Monday, December 19, 2005 5:28 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> Slightly different use case - where you cant just unload and reload
> because of the bug leading to side effect/hang/crash [you have a typo
> or a silly mistake like assigning instead of comparing or you set the
> boolean condition wrong in your if check or your dispatch routine
> indices are all off by one or your variable was left uninitialized and
> you want to initialize before you forget etc] that becomes apparent
> when execution is about to/already hit the code (probably a new change).
> Rebooting seems like an overkill for this.
>
> From Drew’s response, sounds like it is too complex to implement a basic
> one for drivers.
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of
> xxxxx@seagate.com
> Sent: Monday, December 19, 2005 2:56 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> So what stop’s you from doing an Enable/Disable in the Device Manager or
> via DevCon, or your own utility to stop/start your driver?
>
> Gary G. Little
>
>
> —
> You are currently subscribed to windbg as: unknown lmsubst tag argument:
> ‘’
> To unsubscribe send a blank email to xxxxx@lists.osr.com
>
> NOD32 1.1328 (20051219) Information
>
> This message was checked by NOD32 antivirus system.
> http://www.eset.com
>
>
>
>
Hell, David, when you and I started, if you didn’t use assembly you didn’t
program. 
COBOL was the dominant “compiled” language with FORTRAN just
beginning to takeover.
Gary
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of David J. Craig
Sent: Tuesday, December 20, 2005 12:17 AM
To: Kernel Debugging Interest List
Subject: Re:[windbg] Edit and continue
Have you noticed that Windbg has a separate disassembly window? What was
wrong with the interlaced source and assembly window? I don’t like to use
the disassembly windows since it is more difficult to keep the source and
binary together. I will use it and can handle doing my machine code
changes, but most of the newer programmers & engineers seem to avoid
assembly. I started with assembly and no one seems to do that any more. It
is all 4GL or Python or Perl or Forth or xxx.
“Gary G. Little” wrote in message news:xxxxx@windbg…
> Which is a good excuse to learn assembly, as has been pointed out, and
> patch
> it right there in memory to test your theory and then change the source
> and
> rebuild.
>
> Gary G. Little
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
> Sent: Monday, December 19, 2005 5:28 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> Slightly different use case - where you cant just unload and reload
> because of the bug leading to side effect/hang/crash [you have a typo
> or a silly mistake like assigning instead of comparing or you set the
> boolean condition wrong in your if check or your dispatch routine
> indices are all off by one or your variable was left uninitialized and
> you want to initialize before you forget etc] that becomes apparent
> when execution is about to/already hit the code (probably a new change).
> Rebooting seems like an overkill for this.
>
> From Drew’s response, sounds like it is too complex to implement a basic
> one for drivers.
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of
> xxxxx@seagate.com
> Sent: Monday, December 19, 2005 2:56 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> So what stop’s you from doing an Enable/Disable in the Device Manager or
> via DevCon, or your own utility to stop/start your driver?
>
> Gary G. Little
>
>
> —
> You are currently subscribed to windbg as: unknown lmsubst tag argument:
> ‘’
> To unsubscribe send a blank email to xxxxx@lists.osr.com
>
> NOD32 1.1328 (20051219) Information
>
> This message was checked by NOD32 antivirus system.
> http://www.eset.com
>
>
>
>
—
You are currently subscribed to windbg as: glittle@mn.rr.com
To unsubscribe send a blank email to xxxxx@lists.osr.com
NOD32 1.1328 (20051219) Information
This message was checked by NOD32 antivirus system.
http://www.eset.com
Don’t forget about Jovial. Luckily when I started the system was so small
(32K) and the Cobol runtime was 16K, that when processing 24K records from
tape only assembly could be used. It even had a drum. Honeywell 201
minicomputer. I did get to see a Multacs in Phoenix when I went to an
Easycoder class.
“Gary G. Little” wrote in message news:xxxxx@windbg…
Hell, David, when you and I started, if you didn’t use assembly you didn’t
program. COBOL was the dominant “compiled” language with FORTRAN just
beginning to takeover.
Gary
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of David J. Craig
Sent: Tuesday, December 20, 2005 12:17 AM
To: Kernel Debugging Interest List
Subject: Re:[windbg] Edit and continue
Have you noticed that Windbg has a separate disassembly window? What was
wrong with the interlaced source and assembly window? I don’t like to use
the disassembly windows since it is more difficult to keep the source and
binary together. I will use it and can handle doing my machine code
changes, but most of the newer programmers & engineers seem to avoid
assembly. I started with assembly and no one seems to do that any more. It
is all 4GL or Python or Perl or Forth or xxx.
“Gary G. Little” wrote in message news:xxxxx@windbg…
> Which is a good excuse to learn assembly, as has been pointed out, and
> patch
> it right there in memory to test your theory and then change the source
> and
> rebuild.
>
> Gary G. Little
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
> Sent: Monday, December 19, 2005 5:28 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> Slightly different use case - where you cant just unload and reload
> because of the bug leading to side effect/hang/crash [you have a typo
> or a silly mistake like assigning instead of comparing or you set the
> boolean condition wrong in your if check or your dispatch routine
> indices are all off by one or your variable was left uninitialized and
> you want to initialize before you forget etc] that becomes apparent
> when execution is about to/already hit the code (probably a new change).
> Rebooting seems like an overkill for this.
>
> From Drew’s response, sounds like it is too complex to implement a basic
> one for drivers.
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of
> xxxxx@seagate.com
> Sent: Monday, December 19, 2005 2:56 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> So what stop’s you from doing an Enable/Disable in the Device Manager or
> via DevCon, or your own utility to stop/start your driver?
>
> Gary G. Little
>
>
> —
> You are currently subscribed to windbg as: unknown lmsubst tag argument:
> ‘’
> To unsubscribe send a blank email to xxxxx@lists.osr.com
>
> NOD32 1.1328 (20051219) Information
>
> This message was checked by NOD32 antivirus system.
> http://www.eset.com
>
>
>
>
—
You are currently subscribed to windbg as: glittle@mn.rr.com
To unsubscribe send a blank email to xxxxx@lists.osr.com
NOD32 1.1328 (20051219) Information
This message was checked by NOD32 antivirus system.
http://www.eset.com
I couldn’t resist…
We should ask drivers written in MONAD !!!
Robin
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of David J. Craig
Sent: Dienstag, 20. Dezember 2005 18:15
To: Kernel Debugging Interest List
Subject: Re:[windbg] Edit and continue
Don’t forget about Jovial. Luckily when I started the system was so
small
(32K) and the Cobol runtime was 16K, that when processing 24K records
from
tape only assembly could be used. It even had a drum. Honeywell 201
minicomputer. I did get to see a Multacs in Phoenix when I went to an
Easycoder class.
“Gary G. Little” wrote in message
news:xxxxx@windbg…
Hell, David, when you and I started, if you didn’t use assembly you
didn’t
program. COBOL was the dominant “compiled” language with FORTRAN just
beginning to takeover.
Gary
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of David J. Craig
Sent: Tuesday, December 20, 2005 12:17 AM
To: Kernel Debugging Interest List
Subject: Re:[windbg] Edit and continue
Have you noticed that Windbg has a separate disassembly window? What
was
wrong with the interlaced source and assembly window? I don’t like to
use
the disassembly windows since it is more difficult to keep the source
and
binary together. I will use it and can handle doing my machine code
changes, but most of the newer programmers & engineers seem to avoid
assembly. I started with assembly and no one seems to do that any more.
It
is all 4GL or Python or Perl or Forth or xxx.
“Gary G. Little” wrote in message
news:xxxxx@windbg…
> Which is a good excuse to learn assembly, as has been pointed out, and
> patch
> it right there in memory to test your theory and then change the
source
> and
> rebuild.
>
> Gary G. Little
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of Satya Das
> Sent: Monday, December 19, 2005 5:28 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> Slightly different use case - where you cant just unload and reload
> because of the bug leading to side effect/hang/crash [you have a typo
> or a silly mistake like assigning instead of comparing or you set the
> boolean condition wrong in your if check or your dispatch routine
> indices are all off by one or your variable was left uninitialized and
> you want to initialize before you forget etc] that becomes apparent
> when execution is about to/already hit the code (probably a new
change).
> Rebooting seems like an overkill for this.
>
> From Drew’s response, sounds like it is too complex to implement a
basic
> one for drivers.
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of
> xxxxx@seagate.com
> Sent: Monday, December 19, 2005 2:56 PM
> To: Kernel Debugging Interest List
> Subject: RE: [windbg] Edit and continue
>
> So what stop’s you from doing an Enable/Disable in the Device Manager
or
> via DevCon, or your own utility to stop/start your driver?
>
> Gary G. Little
>
>
> —
> You are currently subscribed to windbg as: unknown lmsubst tag
argument:
> ‘’
> To unsubscribe send a blank email to xxxxx@lists.osr.com
>
> NOD32 1.1328 (20051219) Information
>
> This message was checked by NOD32 antivirus system.
> http://www.eset.com
>
>
>
>
—
You are currently subscribed to windbg as: glittle@mn.rr.com
To unsubscribe send a blank email to xxxxx@lists.osr.com
NOD32 1.1328 (20051219) Information
This message was checked by NOD32 antivirus system.
http://www.eset.com
—
You are currently subscribed to windbg as: xxxxx@exgate.tek.com
To unsubscribe send a blank email to xxxxx@lists.osr.com
David J. Craig wrote:
Just think about how you would do something like this. You need to find all
references to that driver. From IoManager and all filters in whatever stack
is involved. Then you have a new driver with code that needs to replace it
while preserving all data changes from the current instance. If the change
was to the size or data order of the device extension, how could it be done?
Doing this to a running OS boggles my mind. It might could be done if
someone had complete control and understanding over all of Windows.
Actually, it isn’t as hard as you might think. “Edit and Continue” is
designed to allow you to change one or two lines. The principle in
Visual Studio is that each function has a bunch of padding at the end.
When you “Edit and Continue”, that one function is recompiled, relinked,
and replaced into memory. If the function grows beyond the padding, it
isn’t possible to “Edit and Continue”.
Other filters and references wouldn’t have to know; no functions are
harmed or moved in the process.
There is no fundamental architectural reason why Windbg couldn’t do this
in the kernel. However, as I said, this would require a much tighter
integration between Windbg and Visual Studio. In my own personal
opinion, the degree of separation we have now is quite healthy.
Just
thinking about this give me a headache, because I don’t think anyone can
truly understand Windows down to the smallest details in all its subsystems.
DOS - yes. Windows - please shoot me.
Yes. In the Windows 3.x days, I felt that I truly understood Windows.
It was possible to trace through the major operations with wdeb386 and
really grok the goings-on. That hasn’t been possible for a decade.
–
Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.
Umm… what makes you think I have not ? In any case, you are talking
about workarounds. Not having EnD is not a show stopper, if it was
there, it would IMO increase the speed at which you do stuff.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Gary G. Little
Sent: Monday, December 19, 2005 8:34 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
Which is a good excuse to learn assembly, as has been pointed out, and
patch it right there in memory to test your theory and then change the
source and rebuild.
Gary G. Little
There will be cases when this will be problematic. It is not an
insurmountable problem however. Remember all the times you could patch
and go ? The debugger could have done that for you. It is as simple as
that.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of David J. Craig
Sent: Monday, December 19, 2005 8:21 PM
To: Kernel Debugging Interest List
Subject: Re:[windbg] Edit and continue
Just think about how you would do something like this. You need to find
all references to that driver. From IoManager and all filters in
whatever stack is involved. Then you have a new driver with code that
needs to replace it while preserving all data changes from the current
instance. If the change was to the size or data order of the device
extension, how could it be done?
Doing this to a running OS boggles my mind. It might could be done if
someone had complete control and understanding over all of Windows.
Just thinking about this give me a headache, because I don’t think
anyone can truly understand Windows down to the smallest details in all
its subsystems.
DOS - yes. Windows - please shoot me.
It is bad enough when the debugger completely owns the address space of
the target application. There the compiler also has the control over
how the code segments are done so a change can be replaced without
impacting areas outside of that process. I suspect that EnC would have
problems with application that establish system hooks and especially
those who load into other processes.
“Satya Das” wrote in message news:xxxxx@windbg…
I did not mean a full IDE but more like a smarter kdfiles that can
refresh code while debugging.
Thanks
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Drew Bliss
Sent: Monday, December 19, 2005 12:10 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] Edit and continue
windbg doesn’t support editing of any kind, so you’re asking windbg to
turn into a full IDE. That is a huge change and not something we’re
planning to do.
—
You are currently subscribed to windbg as: xxxxx@appstream.com To
unsubscribe send a blank email to xxxxx@lists.osr.com
When you say Visual Studio you perhaps mean wdk build ? IMHO it is not
important to have the debugger build the driver. More important is given
a built driver, whether it can update the loaded driver.
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Tim Roberts
Sent: Tuesday, December 20, 2005 10:29 AM
To: Kernel Debugging Interest List
Subject: Re: [windbg] Edit and continue
There is no fundamental architectural reason why Windbg couldn’t do this
in the kernel. However, as I said, this would require a much tighter
integration between Windbg and Visual Studio. In my own personal
opinion, the degree of separation we have now is quite healthy.