Yes I ran ‘make bootstrap’ only. And then I ran it again; it gave
correct output.

Here is some analysis that I could do. Attached are the two log
files. I have kept meaningful names. The ‘build_clean.log’ was the
output after running ‘make bootstrap’ again immediately after the
erroneous build.

1. If you notice at the beginning of ‘build_error.log’ the cleanup ‘rm’
commands are missing; even though I used ‘make bootstrap’. And these
‘rm’ commands were run after the first execution of ‘configure’ (in case
of erroneous build). Second execution of ‘configure’ did not happen in
case of successful build (when I did ‘make bootstrap’ again -
build_clean.log).

2. To my surprise, ‘make -C ../admin/charsets all’ outputs “Nothing to
be done” in case of successful build. I thought ‘make bootstraps’
deletes everything and then starts configure/make. In case of erroneous
build, make in charsets did generate each map.

3. Same was the case with ‘make -C ../admin/grammers’ and a few other
directories.

4. Lots of charset reloading seen in ‘build_error.log’. That did not
happen in  ‘build_clean.log’.

5. First ELC command, after emacs binary is built, produces
error. Perhaps because the binary and pdmp fingerprint mismatch.

One question: is the order of execution deterministic?

Also, the build is now working, should I close this bug?