Development with Eclipse - Wrong error messages in IDE

Peter Moser wrote:

I have some strange error message inside Eclipse, that some symbols cannot
be found. I work with version 9.6 currently. For instance,

Symbol 'RM_HEAP_ID' could not be resolved
src/backend/access/heap/heapam.c

It affects all occurrences of symbols that are defined in
src/include/access/rmgrlist.h. Eclipse just says "Syntax error" here.

However, the source code compiles and runs without any compile-time error or
warning. It is just an IDE problem I think, but it distracts me from finding
real bugs.

Disclaimer: I've never used eclipse.

The problem is some perhaps-too-clever stuff we do to avoid repetitive
declarations of things. The rmgr stuff uses a PG_RMGR macro, which is
defined differently in src/backend/access/transam/rmgr.c and
src/include/access/rmgr.h; the latter contains the actual enum
definition. On the other hand Eclipse is trying to be too clever by
processing the C files, but not actually getting it completely right
(which is understandable, really). We have other similar cases, such as
grammar keywords (kwlist.h)

I'm afraid that you'd have to teach Eclipse to deal with such things
(which might be tricky) or live with it.

--
�lvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Peter Moser wrote:

I have some strange error message inside Eclipse, that some symbols cannot
be found. I work with version 9.6 currently. For instance,

Symbol 'RM_HEAP_ID' could not be resolved
src/backend/access/heap/heapam.c

It affects all occurrences of symbols that are defined in
src/include/access/rmgrlist.h. Eclipse just says "Syntax error" here.

However, the source code compiles and runs without any compile-time error or
warning. It is just an IDE problem I think, but it distracts me from finding
real bugs.

Disclaimer: I've never used eclipse.

The problem is some perhaps-too-clever stuff we do to avoid repetitive
declarations of things. The rmgr stuff uses a PG_RMGR macro, which is
defined differently in src/backend/access/transam/rmgr.c and
src/include/access/rmgr.h; the latter contains the actual enum
definition. On the other hand Eclipse is trying to be too clever by
processing the C files, but not actually getting it completely right
(which is understandable, really). We have other similar cases, such as
grammar keywords (kwlist.h)

I'm afraid that you'd have to teach Eclipse to deal with such things
(which might be tricky) or live with it.

Ok,
thank you for the comment.

I think, I can live with it.

Perhaps, when I have some spare time I give it a try to solve this
"non-issue"...

Cheers,
Peter

--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

On Feb 3, 2016, at 2:38 AM, Peter Moser <pitiz29a@gmail.com> wrote:

Does anyone had similar problems? Do I have to configure Eclipse to understand the PG_RMGR macro or is there another possibility to teach Eclipse these macros?

I just built 9.6 under Eclipse CDT to try this out and was able to open e.g. heapam.c without any error markers.

I added PostgreSQL as a “Makefile Project with Existing Code” after running ./configure from the command-line. After that, I built the project from within Eclipse by adding the ‘all’ make target and running it.

One setting I usually change: right-click the project, pick Properties, then drill down through C/C++ General -> Preprocessor Include Paths. In the Provider pane, there is an entry for “CDT GCC Build Output Parser”. I’m not sure if this is strictly necessary, but I set my “Container to keep discovered entries” setting to “File”.

Basically, Eclipse scans the make output for -I flags, then notes all the includes used to build each file, so the static analyzer, etc. can have more accurate information (it is crucial that the “Compiler command pattern” in this window be a regex that will match the compiler binary you use, so if you have /usr/local/bin/gcc, and “gcc” is the pattern, you are in for trouble).

After running the build, Eclipse should now know what includes are used for each file and stop whining. If it ever seems to have problems, you can kick it by running a clean target, then all, then picking “Project -> C/C++ Index -> Rebuild” (I think).

--
Jason Petersen
Software Engineer | Citus Data
303.736.9255
jason@citusdata.com

On 05.02.2016 um 18:40 Jason Petersen wrote:

On Feb 3, 2016, at 2:38 AM, Peter Moser <pitiz29a@gmail.com> wrote:

Does anyone had similar problems? Do I have to configure Eclipse to understand the PG_RMGR macro or is there another possibility to teach Eclipse these macros?

Hi,

I just built 9.6 under Eclipse CDT to try this out and was able to open e.g. heapam.c without any error markers.

I added PostgreSQL as a “Makefile Project with Existing Code” after running ./configure from the command-line. After that, I built the project from within Eclipse by adding the ‘all’ make target and running it.

I imported PG the same way, configured from terminal with

export CFLAGS="-g0"

./configure \
--prefix="/home/p/pg/build" \
--enable-debug \
--enable-depend \
--enable-cassert

I built the project from command-line, not from within Eclipse. First I
thought that this may have caused all this problems, but no...

One setting I usually change: right-click the project, pick Properties, then drill down through C/C++ General -> Preprocessor Include Paths. In the Provider pane, there is an entry for “CDT GCC Build Output Parser”. I’m not sure if this is strictly necessary, but I set my “Container to keep discovered entries” setting to “File”.

Basically, Eclipse scans the make output for -I flags, then notes all the includes used to build each file, so the static analyzer, etc. can have more accurate information (it is crucial that the “Compiler command pattern” in this window be a regex that will match the compiler binary you use, so if you have /usr/local/bin/gcc, and “gcc” is the pattern, you are in for trouble).

After running the build, Eclipse should now know what includes are used for each file and stop whining. If it ever seems to have problems, you can kick it by running a clean target, then all, then picking “Project -> C/C++ Index -> Rebuild” (I think).

Thanks for all your suggestions, tried all of them, but it made no
difference. What finally solved my problems was to delete the BUILD and
cluster DATA folders that I had created, re-configured the system.
Deleted the project in Eclipse, and imported it again after configure.
All wrong markers disappeared. Maybe the C-file index of Eclipse was
corrupted or so. Also rebuilding it didn't work. The only difference
from my previous attempt was the CFLAGS thing (see above), but I do not
know if this changes anything related to markers.

----
Peter

--
Jason Petersen
Software Engineer | Citus Data
303.736.9255
jason@citusdata.com

--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers