Обсуждение: Subtle bug in autoconf flex version test

The PGAC_PATH_FLEX version test in config/programs.m4 tests the major and minor
versions with = rather than == which unless I’m missing something is performing
assignment rather than testing equality?  The attached diff makes the test
trigger the expected warning on major/minor version on my OS X box (awk version
20070501).

cheers ./daniel

Daniel Gustafsson <daniel@yesql.se> writes:
> The PGAC_PATH_FLEX version test in config/programs.m4 tests the major and minor
> versions with = rather than == which unless I’m missing something is performing
> assignment rather than testing equality?

Huh.  That's been broken since forever ... thanks for noticing!

> The attached diff makes the test
> trigger the expected warning on major/minor version on my OS X box (awk version
> 20070501).

Hm, is that a popular flex version?  I wonder whether we will get
complaints if we start warning about it.
        regards, tom lane

> On 02 May 2016, at 15:38, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Daniel Gustafsson <daniel@yesql.se> writes:
>> The PGAC_PATH_FLEX version test in config/programs.m4 tests the major and minor
>> versions with = rather than == which unless I’m missing something is performing
>> assignment rather than testing equality?
>
> Huh.  That's been broken since forever ... thanks for noticing!
>
>> The attached diff makes the test
>> trigger the expected warning on major/minor version on my OS X box (awk version
>> 20070501).
>
> Hm, is that a popular flex version?  I wonder whether we will get
> complaints if we start warning about it.

Sorry, I missed half the sentence there.  What I meant was that I can trigger
the warning synthetically by changing the version number just to test the
warning; before any version is happily accepted.

cheers ./daniel

Daniel Gustafsson <daniel@yesql.se> writes:
>> On 02 May 2016, at 15:38, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Hm, is that a popular flex version?  I wonder whether we will get
>> complaints if we start warning about it.

> Sorry, I missed half the sentence there.  What I meant was that I can trigger
> the warning synthetically by changing the version number just to test the
> warning; before any version is happily accepted.

Ah, I see.  I was wondering where you found a flex reporting such a
number; I was guessing it was an unreleased devel version ...

Will commit the fix in a bit.
        regards, tom lane