BUG #3909: src\tools\msvc\clean.bat clears parse.h file

On Mon, Jan 28, 2008 at 06:27:05PM +0000, Pavel Golub wrote:

The following bug has been logged online:

Bug reference: 3909
Logged by: Pavel Golub
Email address: pavel@gf.microolap.com
PostgreSQL version: 8.3RC2
Operating system: WinXP
Description: src\tools\msvc\clean.bat clears parse.h file
Details:

Executing clean.bat from src\tools\msvc clears src\backend\parser\parse.h

However, in help stated:

"Bison and Flex are required to build from CVS, but not required when
building from a release file."

Also in src\backend\parser\Makefile we have comments:

"# gram.c, parse.h, and scan.c are in the distribution tarball, so they
# are not cleaned here."

Because of this bug there is no opportunity to build PostgreSQL by Visual
Studio without using Bison.

So, may be as a workaround we may use such scheme:
1. We have "src\backend\parser\parse.h.release" (or whatever postfix) file
2. GenerateFiles() in Solution.pm must check if Bison is available. If so
then generate parse.h by Bison, else copy parse.h.release as parse.h
3. When clean will be called parse.h will be deleted, but not
parse.h.release

I think a better solution is to add a parameter to clean.bat to make it
work like "make clean" does. So you'd to "clean" when you mean "make
clean", and "clean dist" when you mean "make distclean".

Thoughts on this?

//Magnus

On Feb 5, 2008 3:24 PM, Magnus Hagander <magnus@hagander.net> wrote:

On Mon, Jan 28, 2008 at 06:27:05PM +0000, Pavel Golub wrote:

I think a better solution is to add a parameter to clean.bat to make it
work like "make clean" does. So you'd to "clean" when you mean "make
clean", and "clean dist" when you mean "make distclean".

Thoughts on this?

Pretty sure I griped at you before about this, because when it removes
it, it fails to rebuild it the next time round, even on in an env like
mine which can build from CVS perfectly well. More than once I've had
to unpack the tarball again having run a clean.

/D

Dave Page wrote:

On Feb 5, 2008 3:24 PM, Magnus Hagander <magnus@hagander.net> wrote:

On Mon, Jan 28, 2008 at 06:27:05PM +0000, Pavel Golub wrote:
I think a better solution is to add a parameter to clean.bat to make it
work like "make clean" does. So you'd to "clean" when you mean "make
clean", and "clean dist" when you mean "make distclean".

Thoughts on this?

Pretty sure I griped at you before about this, because when it removes
it, it fails to rebuild it the next time round, even on in an env like
mine which can build from CVS perfectly well. More than once I've had
to unpack the tarball again having run a clean.

Uh, that would be a different issue, and I thought that one was fixed. I
certainly clean/rebuild a lot, and it works just fine with the stuff
coming out of cvs.

The problem comes from if you remove *one* of the files but not *all* of
them. We only trigger on one of them (unlike the makefile which triggers
on all). But as long as they are all removed, it should be ok.

Can you confirm if you actually still have that problem with 8.3.0?

//Magnus

On Feb 5, 2008 5:56 PM, Magnus Hagander <magnus@hagander.net> wrote:

Dave Page wrote:

On Feb 5, 2008 3:24 PM, Magnus Hagander <magnus@hagander.net> wrote:

On Mon, Jan 28, 2008 at 06:27:05PM +0000, Pavel Golub wrote:
I think a better solution is to add a parameter to clean.bat to make it
work like "make clean" does. So you'd to "clean" when you mean "make
clean", and "clean dist" when you mean "make distclean".

Thoughts on this?

Pretty sure I griped at you before about this, because when it removes
it, it fails to rebuild it the next time round, even on in an env like
mine which can build from CVS perfectly well. More than once I've had
to unpack the tarball again having run a clean.

Uh, that would be a different issue, and I thought that one was fixed. I
certainly clean/rebuild a lot, and it works just fine with the stuff
coming out of cvs.

The problem comes from if you remove *one* of the files but not *all* of
them. We only trigger on one of them (unlike the makefile which triggers
on all). But as long as they are all removed, it should be ok.

Can you confirm if you actually still have that problem with 8.3.0?

Yup - run clean.bat, and then build.bat and the resulting build fails
because parser/parse.h is missing. Thats with the 8.3.0 tarball, which
built just fine prior to running clean. Bison and flex are in the
path.

/D

On Tue, Feb 05, 2008 at 09:47:16PM +0000, Dave Page wrote:

On Feb 5, 2008 5:56 PM, Magnus Hagander <magnus@hagander.net> wrote:

Dave Page wrote:

On Feb 5, 2008 3:24 PM, Magnus Hagander <magnus@hagander.net> wrote:

On Mon, Jan 28, 2008 at 06:27:05PM +0000, Pavel Golub wrote:
I think a better solution is to add a parameter to clean.bat to make it
work like "make clean" does. So you'd to "clean" when you mean "make
clean", and "clean dist" when you mean "make distclean".

Thoughts on this?

Pretty sure I griped at you before about this, because when it removes
it, it fails to rebuild it the next time round, even on in an env like
mine which can build from CVS perfectly well. More than once I've had
to unpack the tarball again having run a clean.

Uh, that would be a different issue, and I thought that one was fixed. I
certainly clean/rebuild a lot, and it works just fine with the stuff
coming out of cvs.

The problem comes from if you remove *one* of the files but not *all* of
them. We only trigger on one of them (unlike the makefile which triggers
on all). But as long as they are all removed, it should be ok.

Can you confirm if you actually still have that problem with 8.3.0?

Yup - run clean.bat, and then build.bat and the resulting build fails
because parser/parse.h is missing. Thats with the 8.3.0 tarball, which
built just fine prior to running clean. Bison and flex are in the
path.

I've managed to reproduce this now, and it's way more broken than I
thought. In fact, clean.bat simply does not work. It seems the batch
processor falls over *badly* when dealing with so many "call" statements. I
put in some "echo 1", "echo 2" etc to see in which order i runs the bat
file. I put them in from top to bottom, but the execution order came out as
0,5,6,7,8,1,5,6,7,8,7,8,2,3,4,5,6,7,8. Which obviously is less than
perfect.

I've redone a clean.bat now that inlines the deletes, and it seems to be
working better. I'll also add the "clean dist" stuff so it's possible to
clean/rebuild without bison/flex, and then commit this.

//Magnus

On Tue, Feb 05, 2008 at 09:47:16PM +0000, Dave Page wrote:

On Feb 5, 2008 5:56 PM, Magnus Hagander <magnus@hagander.net> wrote:

Dave Page wrote:

On Feb 5, 2008 3:24 PM, Magnus Hagander <magnus@hagander.net> wrote:

On Mon, Jan 28, 2008 at 06:27:05PM +0000, Pavel Golub wrote:
I think a better solution is to add a parameter to clean.bat to make it
work like "make clean" does. So you'd to "clean" when you mean "make
clean", and "clean dist" when you mean "make distclean".

Thoughts on this?

Pretty sure I griped at you before about this, because when it removes
it, it fails to rebuild it the next time round, even on in an env like
mine which can build from CVS perfectly well. More than once I've had
to unpack the tarball again having run a clean.

Uh, that would be a different issue, and I thought that one was fixed. I
certainly clean/rebuild a lot, and it works just fine with the stuff
coming out of cvs.

The problem comes from if you remove *one* of the files but not *all* of
them. We only trigger on one of them (unlike the makefile which triggers
on all). But as long as they are all removed, it should be ok.

Can you confirm if you actually still have that problem with 8.3.0?

Yup - run clean.bat, and then build.bat and the resulting build fails
because parser/parse.h is missing. Thats with the 8.3.0 tarball, which
built just fine prior to running clean. Bison and flex are in the
path.

Fixed version applied to HEAD.

//Magnus

Has this been addressed?

---------------------------------------------------------------------------

Pavel Golub wrote:

The following bug has been logged online:

Bug reference: 3909
Logged by: Pavel Golub
Email address: pavel@gf.microolap.com
PostgreSQL version: 8.3RC2
Operating system: WinXP
Description: src\tools\msvc\clean.bat clears parse.h file
Details:

Executing clean.bat from src\tools\msvc clears src\backend\parser\parse.h

However, in help stated:

"Bison and Flex are required to build from CVS, but not required when
building from a release file."

Also in src\backend\parser\Makefile we have comments:

"# gram.c, parse.h, and scan.c are in the distribution tarball, so they
# are not cleaned here."

Because of this bug there is no opportunity to build PostgreSQL by Visual
Studio without using Bison.

So, may be as a workaround we may use such scheme:
1. We have "src\backend\parser\parse.h.release" (or whatever postfix) file
2. GenerateFiles() in Solution.pm must check if Bison is available. If so
then generate parse.h by Bison, else copy parse.h.release as parse.h
3. When clean will be called parse.h will be deleted, but not
parse.h.release

Regards

---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?

http://www.postgresql.org/docs/faq

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://postgres.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Yes.

//Magnus