bug#49452: Confusing behavior with (include) used in file in GUILE_LOAD_PATH

bug#49452: Confusing behavior with (include) used in file in GUILE_LOAD_PATH

[Vijay Marupudi]

Hello all,

It is unclear to me what the intended behavior for (include
"filename.scm") is, so I'm sending an email about this potential bug.

The Local Inclusion docs
<https://www.gnu.org/software/guile/manual/html_node/Local-Inclusion.html>
seem to state that relative paths are found relative to the file that
included them.

> If file-name is a relative path, it is searched for relative to the
> path that contains the file that the include form appears in.

So if I have a file "/libraries/libname/main.scm" than has (include
"./helpers.scm"), then the file "/libraries/libname/helpers.scm"
*should* (I think) be imported.

But this does not seem to work if "/libraries" is in the GUILE_LOAD_PATH
and my current working directory is somewhere else, say "/home/user" and
I'm running "/home/user/program.scm" that imports the (libname main)
library from "/libraries". Then Guile seems to try to include the
"libname/helpers.scm" file from the current directory, which does not
exist.

Conversations with leoprikler in IRC have revealed to me that
call-with-include-port is the function responsible for this behavior
<https://git.savannah.gnu.org/cgit/guile.git/tree/module/ice-9/psyntax.scm#n3231>.
`syntax-source` returns a file path relative to the load path, and
include tries to use that path to open a file relative to the current
working directory.

In Guile's bug guidelines
<https://www.gnu.org/software/guile/docs/docs-2.2/guile-ref/Reporting-Bugs.html>,
to me this fits

* Whenever documentation and actual behavior differ, you have certainly
  found a bug, either in the documentation or in the program.

and potentially

* When some part of the documentation is not clear and does not make
  sense to you even after re-reading the section, it is a bug.

I believe this is a bug, but I may be wrong, so emailing to clarify.
Thank you!

Vijay Marupudi
PhD Student in Human Centered-Computing
Georgia Institute of Technology

bug#49452: Confusing behavior with (include) used in file in GUILE_LOAD_PATH

[Taylan Kammer]

On 07.07.2021 02:31, Vijay Marupudi wrote:
> Hello all,
> 
> It is unclear to me what the intended behavior for (include
> "filename.scm") is, so I'm sending an email about this potential bug.
> 
> The Local Inclusion docs
> <https://www.gnu.org/software/guile/manual/html_node/Local-Inclusion.html>
> seem to state that relative paths are found relative to the file that
> included them.
> 
>> If file-name is a relative path, it is searched for relative to the
>> path that contains the file that the include form appears in.
> 
> So if I have a file "/libraries/libname/main.scm" than has (include
> "./helpers.scm"), then the file "/libraries/libname/helpers.scm"
> *should* (I think) be imported.
> 
> But this does not seem to work if "/libraries" is in the GUILE_LOAD_PATH
> and my current working directory is somewhere else, say "/home/user" and
> I'm running "/home/user/program.scm" that imports the (libname main)
> library from "/libraries". Then Guile seems to try to include the
> "libname/helpers.scm" file from the current directory, which does not
> exist.
> 
> Conversations with leoprikler in IRC have revealed to me that
> call-with-include-port is the function responsible for this behavior
> <https://git.savannah.gnu.org/cgit/guile.git/tree/module/ice-9/psyntax.scm#n3231>.
> `syntax-source` returns a file path relative to the load path, and
> include tries to use that path to open a file relative to the current
> working directory.
> 
> In Guile's bug guidelines
> <https://www.gnu.org/software/guile/docs/docs-2.2/guile-ref/Reporting-Bugs.html>,
> to me this fits
> 
> * Whenever documentation and actual behavior differ, you have certainly
>   found a bug, either in the documentation or in the program.
> 
> and potentially
> 
> * When some part of the documentation is not clear and does not make
>   sense to you even after re-reading the section, it is a bug.
> 
> I believe this is a bug, but I may be wrong, so emailing to clarify.
> Thank you!
> 
> Vijay Marupudi
> PhD Student in Human Centered-Computing
> Georgia Institute of Technology

Hi Vijay,

I believe this is the same bug as this one I reported 5-6 years ago:

https://bugs.gnu.org/21613

Sadly there was no progress on it as far as I know.

-- 
Taylan

bug#49452: Confusing behavior with (include) used in file in GUILE_LOAD_PATH

[Bengt Richter]

Hi Vijay,

On +2021-07-06 19:31:38 -0500, Vijay Marupudi wrote:
> Hello all,
> 
> It is unclear to me what the intended behavior for (include
> "filename.scm") is, so I'm sending an email about this potential bug.
> 
> The Local Inclusion docs
> <https://www.gnu.org/software/guile/manual/html_node/Local-Inclusion.html>
> seem to state that relative paths are found relative to the file that
> included them.
> 
> > If file-name is a relative path, it is searched for relative to the
> > path that contains the file that the include form appears in.
> 
> So if I have a file "/libraries/libname/main.scm" than has (include
> "./helpers.scm"), then the file "/libraries/libname/helpers.scm"
> *should* (I think) be imported.
>

A little nit, but
    (include "filename.scm")
is not identical to
    (include "./filename.scm")
so might it be worth trying
   (include "helpers.scm")
in your code?

And/or should the docs explain that both ways of writing a relative
file name in this context are or are not acceptable?

Might some macro dislike leading dots?

Long shot ;-)

> But this does not seem to work if "/libraries" is in the GUILE_LOAD_PATH
> and my current working directory is somewhere else, say "/home/user" and
> I'm running "/home/user/program.scm" that imports the (libname main)
> library from "/libraries". Then Guile seems to try to include the
> "libname/helpers.scm" file from the current directory, which does not
> exist.
> 
> Conversations with leoprikler in IRC have revealed to me that
> call-with-include-port is the function responsible for this behavior
> <https://git.savannah.gnu.org/cgit/guile.git/tree/module/ice-9/psyntax.scm#n3231>.
> `syntax-source` returns a file path relative to the load path, and
> include tries to use that path to open a file relative to the current
> working directory.
> 
> In Guile's bug guidelines
> <https://www.gnu.org/software/guile/docs/docs-2.2/guile-ref/Reporting-Bugs.html>,
> to me this fits
> 
> * Whenever documentation and actual behavior differ, you have certainly
>   found a bug, either in the documentation or in the program.
> 
> and potentially
> 
> * When some part of the documentation is not clear and does not make
>   sense to you even after re-reading the section, it is a bug.
> 
> I believe this is a bug, but I may be wrong, so emailing to clarify.
> Thank you!
> 
> Vijay Marupudi
> PhD Student in Human Centered-Computing
> Georgia Institute of Technology
> 
> 
> 
> 

-- 
Regards,
Bengt Richter

bug#49452: Confusing behavior with (include) used in file in GUILE_LOAD_PATH

[Vijay Marupudi]

Thank you Taylan. This does appear to be the same bug. This bug can be
closed as a duplicate. Looks like Andy Wingo suggested a potential fix,
but there was no follow-up. If someone is willing to show me the ropes
when I need them, I can take a stab at learning the Guile codebase and
fixing the bug.

Vijay

On 7/7/21 3:29 AM, Taylan Kammer wrote:
> On 07.07.2021 02:31, Vijay Marupudi wrote:
>> Hello all,
>>
>> It is unclear to me what the intended behavior for (include
>> "filename.scm") is, so I'm sending an email about this potential bug.
>>
>> The Local Inclusion docs
>> <https://www.gnu.org/software/guile/manual/html_node/Local-Inclusion.html>
>> seem to state that relative paths are found relative to the file that
>> included them.
>>
>>> If file-name is a relative path, it is searched for relative to the
>>> path that contains the file that the include form appears in.
>>
>> So if I have a file "/libraries/libname/main.scm" than has (include
>> "./helpers.scm"), then the file "/libraries/libname/helpers.scm"
>> *should* (I think) be imported.
>>
>> But this does not seem to work if "/libraries" is in the GUILE_LOAD_PATH
>> and my current working directory is somewhere else, say "/home/user" and
>> I'm running "/home/user/program.scm" that imports the (libname main)
>> library from "/libraries". Then Guile seems to try to include the
>> "libname/helpers.scm" file from the current directory, which does not
>> exist.
>>
>> Conversations with leoprikler in IRC have revealed to me that
>> call-with-include-port is the function responsible for this behavior
>> <https://git.savannah.gnu.org/cgit/guile.git/tree/module/ice-9/psyntax.scm#n3231>.
>> `syntax-source` returns a file path relative to the load path, and
>> include tries to use that path to open a file relative to the current
>> working directory.
>>
>> In Guile's bug guidelines
>> <https://www.gnu.org/software/guile/docs/docs-2.2/guile-ref/Reporting-Bugs.html>,
>> to me this fits
>>
>> * Whenever documentation and actual behavior differ, you have certainly
>>   found a bug, either in the documentation or in the program.
>>
>> and potentially
>>
>> * When some part of the documentation is not clear and does not make
>>   sense to you even after re-reading the section, it is a bug.
>>
>> I believe this is a bug, but I may be wrong, so emailing to clarify.
>> Thank you!
>>
>> Vijay Marupudi
>> PhD Student in Human Centered-Computing
>> Georgia Institute of Technology
> 
> Hi Vijay,
> 
> I believe this is the same bug as this one I reported 5-6 years ago:
> 
> https://bugs.gnu.org/21613
> 
> Sadly there was no progress on it as far as I know.
> 

bug#49452: Confusing behavior with (include) used in file in GUILE_LOAD_PATH

[Vijay Marupudi]

Hello Bengt,

Unfortunately that still doesn't work :( I think they should be
considered the same though, when this bug gets fixed.

Vijay

On 7/7/21 4:22 PM, Bengt Richter wrote:
> Hi Vijay,
> 
> On +2021-07-06 19:31:38 -0500, Vijay Marupudi wrote:
>> Hello all,
>>
>> It is unclear to me what the intended behavior for (include
>> "filename.scm") is, so I'm sending an email about this potential bug.
>>
>> The Local Inclusion docs
>> <https://www.gnu.org/software/guile/manual/html_node/Local-Inclusion.html>
>> seem to state that relative paths are found relative to the file that
>> included them.
>>
>>> If file-name is a relative path, it is searched for relative to the
>>> path that contains the file that the include form appears in.
>>
>> So if I have a file "/libraries/libname/main.scm" than has (include
>> "./helpers.scm"), then the file "/libraries/libname/helpers.scm"
>> *should* (I think) be imported.
>>
> 
> A little nit, but
>     (include "filename.scm")
> is not identical to
>     (include "./filename.scm")
> so might it be worth trying
>    (include "helpers.scm")
> in your code?
> 
> And/or should the docs explain that both ways of writing a relative
> file name in this context are or are not acceptable?
> 
> Might some macro dislike leading dots?
> 
> Long shot ;-)
> 
>> But this does not seem to work if "/libraries" is in the GUILE_LOAD_PATH
>> and my current working directory is somewhere else, say "/home/user" and
>> I'm running "/home/user/program.scm" that imports the (libname main)
>> library from "/libraries". Then Guile seems to try to include the
>> "libname/helpers.scm" file from the current directory, which does not
>> exist.
>>
>> Conversations with leoprikler in IRC have revealed to me that
>> call-with-include-port is the function responsible for this behavior
>> <https://git.savannah.gnu.org/cgit/guile.git/tree/module/ice-9/psyntax.scm#n3231>.
>> `syntax-source` returns a file path relative to the load path, and
>> include tries to use that path to open a file relative to the current
>> working directory.
>>
>> In Guile's bug guidelines
>> <https://www.gnu.org/software/guile/docs/docs-2.2/guile-ref/Reporting-Bugs.html>,
>> to me this fits
>>
>> * Whenever documentation and actual behavior differ, you have certainly
>>   found a bug, either in the documentation or in the program.
>>
>> and potentially
>>
>> * When some part of the documentation is not clear and does not make
>>   sense to you even after re-reading the section, it is a bug.
>>
>> I believe this is a bug, but I may be wrong, so emailing to clarify.
>> Thank you!
>>
>> Vijay Marupudi
>> PhD Student in Human Centered-Computing
>> Georgia Institute of Technology
>>
>>
>>
>>
>