Re: 05no-escape-on-AT_CHECK-strings

[Akim Demaille]

| Akim Demaille wrote:
| > Derek> Akim Demaille wrote:
| > >> I am absolutely against this patch.  Revert it.
| > >> >> If you want something like this, you will first need to
| > explain it.
| > >> Derek> Again, CVS's current test suite often compares program
| > output against
| > Derek> shell variables defined earlier in the script.
| >
| >Then save these values in an expout file, and use [expout] as STDOUT
| >value.
| >
| 
| The CVS test suite uses this almost every test.  I could implement
| this as an m4 macro in the CVS test suite, but had thought it had some
| more general usefulness.  How would you feel about an extra macro,
| perhaps AT_CHECK_NOESCAPE or the like which ran AT_CHECK to expout
| then allowed shell expansion on the args before comparison?

This, why not.

| I also wanted to implement an AT_CHECK_EXPR (now AT_CHECK_EXPR &
| AT_CHECK_EXPR_NOESCAPE?) which used expr rather than diff to compare
| the strings. 

expr suffers from severe limitations, including portability issues.

| One example of where this is useful with CVS is in that
| CVS's includes dates in much of its output.  Since these dates change
| depending on when the test is run, CVS's test suite often uses
| ${DATE}, which expands to a regexp which will match the date format
| though no particular date, to match dates in tests where the date is
| not the central issue being tested.

This is where I suggest a wrapper that smashes these differences.