4 3

[alpine-devel] Regex compatibility

Michael Fox
Details
Message ID
<CABbL6oaVAbNBCM_3dfrA5BqTDn-f6XE2vzaDNbg0S5fPkEnYtg@mail.gmail.com>
Sender timestamp
1435622028
DKIM signature
missing
Download raw message
Building software on alpine, I notice most problems stem from regex
incompatibility. Is there some way achieve compatibility?

This is especially hard on `make check` because many tests involve regexes.
To the point that I think a lot of packagers are just not running `make
check` because they figure the tests are failing because of test code and
not the code under test. This may be true most of the time but it is
allowing some bugs to slip through.

-- 

-
Michael
Jeff Pohlmeyer
Details
Message ID
<CAEuRe+3jAhHc5kYuN2VHx6dxe+90Jw4=+wp5UzAns_YTrtMbRw@mail.gmail.com>
In-Reply-To
<CABbL6oaVAbNBCM_3dfrA5BqTDn-f6XE2vzaDNbg0S5fPkEnYtg@mail.gmail.com> (view parent)
Sender timestamp
1435669502
DKIM signature
missing
Download raw message
On Mon, Jun 29, 2015 at 6:53 PM, Michael Fox <415fox@gmail.com> wrote:
> Building software on alpine, I notice most problems stem from regex
> incompatibility. Is there some way achieve compatibility?
>
> This is especially hard on `make check` because many tests involve regexes.

If it's a matter of the tests actually depending on regex library
calls, I would say
that's something that would probably need to be fixed in the test, since not all
C libraries are going to have the same implementation. But if the tests are
depending on the behavior of the "grep" command, the fix might be as simple
as ensuring that the test environment has GNU grep installed in the path ahead
of busybox grep. (since GUN grep uses PCRE rather than MUSL for its regex.)

- Jeff


---
Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
Help:         alpine-devel+help@lists.alpinelinux.org
---
Michael Fox
Details
Message ID
<CABbL6obAHg2_rSzJ=5NKMmKG1eCp3YEdX4e4okXjf-70D4cuVQ@mail.gmail.com>
In-Reply-To
<20150630162542.34f46f2a@ncopa-desktop.alpinelinux.org> (view parent)
Sender timestamp
1435684319
DKIM signature
missing
Download raw message
Here's all the things I've installed to get tests to pass:

gnugrep binutils diffutils coreutils linux-headers gzip bash

And still, of the many basic, core packages I've tested, very few pass all
tests. Maybe less than one in four.

There's currently 7575 packages in alpine. How much work is that to report
failing tests upstream? Even more work to try and make your own patches.

Do you think we can get 200 people to take on a package each day for a
month? If so, I volunteer to be one of them.


On Tue, Jun 30, 2015 at 7:25 AM, Natanael Copa <ncopa@alpinelinux.org>
wrote:

> On Mon, 29 Jun 2015 16:53:48 -0700
> Michael Fox <415fox@gmail.com> wrote:
>
> > Building software on alpine, I notice most problems stem from regex
> > incompatibility. Is there some way achieve compatibility?
>
> yes. use portable regexes and avoid gnu (and other platform) extensions.
>
> > This is especially hard on `make check` because many tests involve
> regexes.
> > To the point that I think a lot of packagers are just not running `make
> > check` because they figure the tests are failing because of test code and
> > not the code under test. This may be true most of the time but it is
> > allowing some bugs to slip through.
>
> buts in 'make check' scripts is likely slippering though yes.
>
> I suppose other option is to install gnu sed and gnu grep when doing
> make check. But for longterm, I'd prefer upstream project think more
> about portability.
>
> -nc
>



-- 

-
Michael
Michael Fox
Details
Message ID
<CABbL6oZU0v_5_vVcTeTKADnVOPCp3jzsHCfzBfLa0hJX=b8Qpg@mail.gmail.com>
In-Reply-To
<CABbL6obAHg2_rSzJ=5NKMmKG1eCp3YEdX4e4okXjf-70D4cuVQ@mail.gmail.com> (view parent)
Sender timestamp
1435687937
DKIM signature
missing
Download raw message
On second thought, `apk search '*'` spits out a bunch of repeats for some
reason. Then there's dev and doc packages. There's only 3364 unique
packages.

On Tue, Jun 30, 2015 at 10:11 AM, Michael Fox <415fox@gmail.com> wrote:

> Here's all the things I've installed to get tests to pass:
>
> gnugrep binutils diffutils coreutils linux-headers gzip bash
>
> And still, of the many basic, core packages I've tested, very few pass all
> tests. Maybe less than one in four.
>
> There's currently 7575 packages in alpine. How much work is that to report
> failing tests upstream? Even more work to try and make your own patches.
>
> Do you think we can get 200 people to take on a package each day for a
> month? If so, I volunteer to be one of them.
>
>
> On Tue, Jun 30, 2015 at 7:25 AM, Natanael Copa <ncopa@alpinelinux.org>
> wrote:
>
>> On Mon, 29 Jun 2015 16:53:48 -0700
>> Michael Fox <415fox@gmail.com> wrote:
>>
>> > Building software on alpine, I notice most problems stem from regex
>> > incompatibility. Is there some way achieve compatibility?
>>
>> yes. use portable regexes and avoid gnu (and other platform) extensions.
>>
>> > This is especially hard on `make check` because many tests involve
>> regexes.
>> > To the point that I think a lot of packagers are just not running `make
>> > check` because they figure the tests are failing because of test code
>> and
>> > not the code under test. This may be true most of the time but it is
>> > allowing some bugs to slip through.
>>
>> buts in 'make check' scripts is likely slippering though yes.
>>
>> I suppose other option is to install gnu sed and gnu grep when doing
>> make check. But for longterm, I'd prefer upstream project think more
>> about portability.
>>
>> -nc
>>
>
>
>
> --
>
> -
> Michael
>



-- 

-
Michael
Natanael Copa
Details
Message ID
<20150630162542.34f46f2a@ncopa-desktop.alpinelinux.org>
In-Reply-To
<CABbL6oaVAbNBCM_3dfrA5BqTDn-f6XE2vzaDNbg0S5fPkEnYtg@mail.gmail.com> (view parent)
Sender timestamp
1435674342
DKIM signature
missing
Download raw message
On Mon, 29 Jun 2015 16:53:48 -0700
Michael Fox <415fox@gmail.com> wrote:

> Building software on alpine, I notice most problems stem from regex
> incompatibility. Is there some way achieve compatibility?

yes. use portable regexes and avoid gnu (and other platform) extensions.

> This is especially hard on `make check` because many tests involve regexes.
> To the point that I think a lot of packagers are just not running `make
> check` because they figure the tests are failing because of test code and
> not the code under test. This may be true most of the time but it is
> allowing some bugs to slip through.

buts in 'make check' scripts is likely slippering though yes.

I suppose other option is to install gnu sed and gnu grep when doing
make check. But for longterm, I'd prefer upstream project think more
about portability.

-nc


---
Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
Help:         alpine-devel+help@lists.alpinelinux.org
---