I don't know if this is the right place to report this...

Unless I'm missing something, GenTable from the topGO package has a serious bug in that the column reporting the results of the test is of type character instead of numeric. Example:

library(topGO)

data(GOdata)
data(results.tGO)
sig.tab <- GenTable(GOdata, Fis = resultFisher, topNodes = 20)

str(sig.tab)
'data.frame':   20 obs. of  6 variables:
 $ GO.ID      : chr  "GO:0006091" "GO:0022900" "GO:0009267" "GO:0042594" ...
 $ Term       : chr  "generation of precursor metabolites and ..." "electron transport chain" "cellular response to starvation" "response to starvation" ...
 $ Annotated  : int  8 8 6 6 30 30 9 9 7 7 ...
 $ Significant: int  6 6 5 5 12 12 6 6 5 5 ...
 $ Expected   : num  1.26 1.26 0.94 0.94 4.72 4.72 1.42 1.42 1.1 1.1 ...
 $ Fis        : chr  "0.00025" "0.00025" "0.00043" "0.00043" ...
                ^^^

This means that results like '2.9e-06' would be filtered out by something like Fis < 0.01

sessionInfo()
R version 4.0.3 (2020-10-10)
Platform: x86_64-conda-linux-gnu (64-bit)
Running under: CentOS Linux 7 (Core)

Matrix products: default
BLAS/LAPACK: /export/projects/III-data/wcmp_bioinformatics/db291g/miniconda3/envs/20210301_nelson_act_mutants/lib/libopenblasp-r0.3.10.so

locale:
 [1] LC_CTYPE=en_GB.UTF-8       LC_NUMERIC=C               LC_TIME=en_GB.UTF-8        LC_COLLATE=en_GB.UTF-8     LC_MONETARY=en_GB.UTF-8    LC_MESSAGES=en_GB.UTF-8    LC_PAPER=en_GB.UTF-8       LC_NAME=C                  LC_ADDRESS=C              
[10] LC_TELEPHONE=C             LC_MEASUREMENT=en_GB.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats4    parallel  stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] topGO_2.42.0         SparseM_1.81         GO.db_3.12.1         AnnotationDbi_1.52.0 IRanges_2.24.0       S4Vectors_0.28.0     Biobase_2.50.0       graph_1.68.0         BiocGenerics_0.36.0 

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.6         matrixStats_0.58.0 lattice_0.20-41    grid_4.0.3         DBI_1.1.1          RSQLite_2.2.3      cachem_1.0.4       rlang_0.4.10       blob_1.2.1         vctrs_0.3.6        tools_4.0.3        bit64_4.0.5        bit_4.0.4          fastmap_1.1.0     
[15] compiler_4.0.3     pkgconfig_2.0.3    memoise_2.0.0

Nope, you aren't missing anything. The GenTable function calls format.FUN which calls format to make the values 'pretty'. And format returns character results.

> format(1:5)
[1] "1" "2" "3" "4" "5"

As to whether or not this constitutes a bug, I imagine that is up to Adrian Alexa? I mean, a bug is when a function returns something that the programmer didn't intend. Perhaps the intention was that the p-values should get run through format in order to make the resulting table pretty, and there was never any intent that the table should be easily subsetted on the p-value. I actually have no idea, and so far as I can tell Adrian doesn't really actively maintain topGO any longer?

$ git log -n 200 | grep Alexa --after-context=2
Author: Adrian Alexa <adrian.alexa@gmail.com>
Date:   Tue Dec 10 01:36:39 2019 +0000

--
Author: Adrian Alexa <adrian.alexa@gmail.com>
Date:   Wed Feb 28 02:08:18 2018 +0000

--
Author: Adrian Alexa <adrian.alexa@gmail.com>
Date:   Wed Feb 28 02:06:18 2018 +0000

--
Author: Adrian Alexa <alexa@mpi-inf.mpg.de>
Date:   Thu Apr 10 15:41:48 2014 +0000

--
Author: Adrian Alexa <alexa@mpi-inf.mpg.de>
Date:   Mon Oct 10 00:07:19 2011 +0000

--
Author: Adrian Alexa <alexa@mpi-inf.mpg.de>
Date:   Mon Oct 10 00:04:58 2011 +0000

--
Author: Adrian Alexa <alexa@mpi-inf.mpg.de>
Date:   Thu Oct 14 17:33:29 2010 +0000

--
Author: Adrian Alexa <alexa@mpi-inf.mpg.de>
Date:   Fri Jun 4 14:28:07 2010 +0000

Or maybe it's fully mature and doesn't need any more work. Anyway, it's easy enough to cast to numeric using as.numeric, so there's that.